Trends in Analytical Chemistry 52 (2013) 261–274

Contents lists available at ScienceDirect

Trends in Analytical Chemistry

journal homepage: www.elsevier .com/locate / t rac

Review

Recent advances in emerging imaging techniques for non-destructive detectionof food quality and safety

Quansheng Chen ⇑, Chaojie Zhang, Jiewen Zhao, Qin OuyangSchool of Food & Biological Engineering, Jiangsu University, Zhenjiang 212013, China

a r t i c l e i n f o

Keywords:Fluorescence imaging

Food qualityFood safetyHyperspectral imagingMagnetic resonance imagingNon-destructive detectionOdor imagingSoft X-ray imagingThermal imagingUltrasound imaging

0165-9936/$ - see front matter � 2013 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.trac.2013.09.007

Abbreviations: ANN, Artificial neural network; Bintegral; BIL, Band interleaved by line; BIP, Band intsequential; CCD, Charge-coupled device; CNR, Contrputed tomography; FI, Fluorescence imaging; FMimaging; FPA, Focal plane array; HCA, HierarchicHyperspectral imaging; IR, Infrared; LCSM, Laser coLDA, Linear discriminant analysis; LED, Light-emittianalysis; MLR, Multi-linear regression; MRI, MagnetNuclear magnetic resonance; OI, Odor imaging; PCA, PQA, Quality assurance; QC, Quality control; VIS/NIRRadial basis function; RF, Radio frequency; RFC, RadiRadio frequency sampling; RMI, Raman microscopic imachine; 3D, Three-dimensional; TI, Thermal imagiVOC, Volatile organic compound; XRI, X-ray imaging.⇑ Corresponding author. Tel.: +86 511 88790318; F

E-mail address: qschen@ujs.edu.cn (Q. Chen).

a b s t r a c t

Food quality and safety issues are increasingly attracting attention. Emerging imaging techniques haveparticular advantages in non-destructive detection of food quality and safety. This review looks at thetrends in applying these emerging imaging techniques to analysis of food quality and safety, in particular,hyperspectral imaging, magnetic resonance imaging, soft X-ray imaging, ultrasound imaging, thermalimaging, fluorescence imaging, and odor imaging. On the basis of the observed trends, we also presentthe technical challenges and future outlook for these emerging imaging techniques.

� 2013 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2612. Emerging imaging techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262

2.1. Hyperspectral imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2622.2. Magnetic resonance imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2652.3. Soft X-ray imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2652.4. Ultrasound imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2672.5. Thermal imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2672.6. Fluorescence imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2682.7. Odor imaging. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269

3. Technical challenges and trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2704. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272

ll rights reserved.

AI, Backscattered amplitudeerleaved by pixel; BSQ, Bandast-to-noise ratio; CT, Com-I, Fluorescence microscopical clustering analysis; HSI,nfocal scanning microscopy;ng diode; MFA, Multifractalic resonance imaging; NMR,rincipal component analysis;, Visible/Near infrared; RBF,o frequency correlation; RFS,maging; SVM, Support vectorng; UI, Ultrasound imaging;

ax: +86 511 88780201.

1. Introduction

Food is any substance consumed to provide nutritional supportfor the human body. It is usually of plant or animal origin, and con-tains essential nutrients, such as carbohydrates, fats, proteins, vita-mins and minerals. Today, most of the food energy consumed bythe world population is supplied by food industry. Food qualityand safety control, which directly relates to human health andthe sustainable development of a country, has received specialemphasis from government and has attracted great social concernand global attention. With the rapid development of economies,consumers’ rising and persistent demand for safe food and betterquality of food and beverage is emphasized. Food quality involvesthe quality characteristics of food that are acceptable to consum-ers, including such external factors as appearance (size, shape, col-

262 Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274

or, gloss and consistency), texture, and flavor, and other internalfactors (chemical, physical and microbial).

Food quality is an important food-manufacturing require-ment, because consumers are susceptible to any form of contam-ination that may occur during the manufacturing processes.Besides ingredient quality control (QC), there are also sanitationrequirements. It is important to ensure that the food-processingenvironment is as clean as possible in order to produce the saf-est possible food for the consumer. Food quality also needs aproduct traceability and recall system. It also deals with labelingissues to ensure there are correct ingredient and nutritionalinformation.

Food safety is a scientific discipline describing handling, prepa-ration, and storage of food in ways that prevent foodborne illness.Food hazard can be divided into three parts:

� physical hazards (staples, nails, screws, bolts, glass particles andsplinters);� chemical hazards (pesticides, mold, herbicides, contamination

from rodents, grease, heavy metals, washing and sanitarycompounds);� environmental pollution and pollution caused by human activ-

ity (pathogenic bacteria from soil, excrements, parasites andviruses).

Pathogenic bacteria can be transferred by means of poor hy-giene, human diseases and contaminated water or compost [1].Food can serve as a growth medium that consists largely of water,protein, lipid and polysaccharides for bacteria that can cause foodpoisoning. Debates on genetic food safety include such issues asimpact of genetically-modified food on the health of further gen-erations and genetic pollution of the environment, which can de-stroy natural biological diversity. It is especially important forfoods and agricultural products to have a high degree of safety,and inspections for QC include biological, chemical and physicaltests.

Food quality assurance (QA) has always been one of the mostdifficult problems associated with handling, processing, sortingand ensuring safety in the food industry. Food quality is moni-tored traditionally by human panel test, chemical analytical mea-surement and mechanical methods. The human panel test hasbeen widely used in food-quality assessment, for example, tea-quality grade identification, wine-quality analysis and dairy-prod-uct evaluation. However, this method is time consuming andpurely subjective, and is affected by external factors (e.g., adapta-tion, fatigue and state of mind) [1]. For the determination of thecomponents related to food quality and safety, conventionalchemical analysis is generally adopted. But this method is expen-sive, laborious, and invasive, and it is possible only in laboratoriessince instruments are required for the purpose. Moreover, compli-cated sample preprocessing is usually required, and causesdifficulty for real-time and on-line monitoring in food manufac-turing. At present, foreign substances in food are detected usingmainly mechanical and optical methods. These techniques detecta large portion of the foreign substances due to their difference inmass (mechanical sieving), color (optical method) and surfacedensity (ultrasonic detection). Despite the numerous differentmethods, a considerable proportion of the foreign substances re-main undetected [2]. Thus, non-destructive, non-contact and fastmeasurement methods are in great demand for on-line industrialQC tasks.

In recent years, imaging technologies have become valuabletools in all major areas of application, particularly due to recenttechnological developments in camera technology and the process-ing power of computer hardware. This review looks at the status inthe application of emerging imaging techniques to detection of

food quality and safety especially reviewing the applications ofhyperspectral imaging (HSI), magnetic resonance imaging (MRI),X-ray imaging (XRI), ultrasound imaging (UI), thermal imaging(TI), fluorescence imaging (FI) and odor imaging (OI). Each of thewide range of different imaging modalities has its own individualcharacteristics. For example, XRI is appropriate if a fracture is sus-pected, because X-rays are good at imaging bones [3]. The relativeadvantage of any particular modality lies essentially in the mech-anism of the contrast in the images that it produces. In this respect,UI is rather versatile [4].

Previous research papers and reviews, complementary to thescope of this review, have covered broader and related areas of re-search. Señorans et al. [5] reviewed a group of new analytical tech-niques, including food image analysis, and their use for food andprocess control. Du et al. [6] summarized recent developments inimage-processing techniques for food-quality evaluation; imageacquisition, image segmentation, feature extraction and classifica-tion methods were reviewed. Mathiassen et al. [7] reviewed imag-ing technologies for inspecting fish and fish products, whichincluded visible/near-infrared light (VIS/NIR) imaging, VIS/NIRspectral imaging, computed tomography (CT) XRI, and MRI.

Herein, we review in detail the trends in the application ofemerging imaging techniques to analysis of food quality and safety,particularly HSI, MRI, soft XRI, UI, TI, FI and OI techniques. We alsopresent the technical challenges and future outlook for theseemerging imaging techniques.

2. Emerging imaging techniques

2.1. Hyperspectral imaging

The HSI technique, as a chemical or spectroscopic imaging ana-lytical tool, has found application in diverse fields, such as astron-omy, agriculture, pharmaceuticals, and medicine. It is an emergingtechnique that integrates conventional imaging and spectroscopyto attain both spatial and spectral information from an object.The images obtained, commonly called hypercubes, are three-dimensional (3D) data cubes, which are made up of hundreds ofcontiguous wavebands for each spatial position of a target studied,as shown in Fig. 1. Consequently, spectra of each pixel can be usedto characterize the composition of that specific position, and sur-face-feature information can be obtained according to the spatialimages. There are two conventional methods for hyperspectral im-age acquisition namely, the ‘‘staring imager’’ configuration and‘‘push-broom’’ acquisition [8].

The ‘‘push-broom’’ acquisition involves acquisition of simulta-neous spectral measurements from a series of adjacent spatialpositions – this requires relative movement between the objectand the detector. Some instruments produce hyperspectral imagesbased on a point step and acquiring mode: spectra are obtained atsingle points on a sample, and then the sample is moved and an-other spectrum taken. Hypercubes obtained using this configura-tion are stored as the band interleaved by pixel (BIP) format.Advances in detector technology have reduced the time requiredto acquire hypercubes. Line-mapping instruments record the spec-trum of each pixel in a line of sample that is simultaneously re-corded by an array detector; and the resultant hypercube isstored in the band interleaved by line (BIL) format [8]. HSI systemsbased on the ‘‘push-broom’’ acquisition typically contain the fol-lowing components: objective lens, spectrograph, camera, acquisi-tion system, translation stage, illumination and computer. Fig. 2(a)shows an HSI system based on a ‘‘push-broom’’ acquisition, and a3D data cube, often called ‘‘hypercube’’, was obtained using thissystem, as shown in Fig. 2(b).

Fig. 1. Hypercube showing the relationship between spectral and spatial dimensions.

Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274 263

However, the hypercube through the ‘‘push-broom’’ acquisi-tion is very huge, and consumes too much time in further dataprocessing; besides, this system is incontestably expensive. Asa result, it is of great interest to develop a low-cost, rapid HSIsystem for detection of food quality and safety. In general, a‘‘push-broom’’ acquisition is, going through spectral preprocess-ing, image processing and model recognition to reach the in-tended goals in laboratory; and some optimum images at thespecific wavebands can then be found from the hypercube. Thus,we select the right filters corresponding to some specific wave-bands to develop a multi-spectral imaging system for the real-time usage.

Conventionally, hyperspectral image acquisition by a multi-spectral imaging system is also called the ‘‘staring imager’’ config-uration, which involves keeping the image field of view fixed, andobtaining images one wavelength after another. Hypercubes ob-tained using this configuration thus comprise a 3D stack of images(one image for each wavelength examined), stored in what isknown as the band sequential (BSQ) format, as shown inFig. 2(d). Wavelength in the ‘‘staring imager’’ configuration is typ-ically moderated using a tunable filter [8]. The simplest method toobtain images at a discrete spectral region is by positioning a band-pass filter (or interference filter) in front of a monochrome cameralens. Hypercubes can be obtained by capturing a series of spectralimages by sequentially changing filters in front of the camera.Fig. 2(c) shows a multi-spectral imaging system based on a rotatingfilter wheel.

Integrating spectroscopy with image analysis, the HSI techniqueshows its superiority to the most commonly-used imaging tech-niques, and it is particularly interesting among the emerging tech-niques applied in food sector [9]. HSI can be carried out inreflectance, transmission or fluorescence modes, although thecurrent published research on HSI is based more on reflectance

mode than on transmission and emission modes. This techniquehas recently emerged as a powerful analytical tool for rapid, non-contact and non-destructive food assessment.

In food-quality inspection, detection of defects on diverse foodshas been investigated. Mehl et al. [10] studied the detection of de-fects on selected apple cultivars using HSI and multi-spectral imag-ing techniques. A multi-spectral imaging system with specificfilters was designed based on spectral features of apples character-ized by HSI analysis, and good isolation of scabs, fungal and soilcontaminations, and bruises was obtained using PCA or the chloro-phyll-absorption peak.

Gowen et al. [11] investigated damage detection on the caps ofwhite mushrooms (Agaricus bisporus) using a push-broom line-scanning HSI instrument, and illustrated the potential of the sys-tem for non-destructive monitoring damaged mushrooms on theprocessing line.

Lu et al. [12] applied VIS/NIR HSI combined with radial basisfunction support vector machine (RBF-SVM) classification to in-spect hidden bruises on kiwi fruits.

Much research has been attempted for the application of HSI tofood-quality classification. ElMasry et al. [13] developed an NIR HSIsystem to assess the quality of cooked turkey hams with differentingredients and processing parameters.

Maftoonazad et al. [14] used the HSI technique to model qualitychanges in avocados during storage at different temperatures, inwhich multilayer artificial neural networks (ANNs) were used todevelop models. This inspection approach based on image analysisand processing also found satisfactory results in tea-quality assess-ment [15–20]. Further, researchers have explored the applicationsof HSI in other food-quality analyses, such as pear [21], citrus fruits[22], tomato [23], and pork [24,25].

Concerning food-safety control, the HSI technique has alsomade a great contribution. Fecal contamination is an important

Fig. 2. Hyperspectral imaging (HSI) system and three-dimensional hypercube. (a) HSI system based on a ‘‘push-broom’’ acquisition. (b) Hypercube by a ‘‘push-broom’’acquisition. (c) HSI system based on a ‘‘staring imager’’ acquisition. (d) Hypercube obtained by a ‘‘staring imager’’ acquisition.

264 Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274

food-safety issue. Kim et al. [26] used a recently-developed HSIsystem to examine experimentally contaminated apples and iden-tify potential wavelengths that could be used in an on-line multi-spectral imaging system.

Yang et al. [27] optimized and evaluated the multi-spectralalgorithms for detection of fecal contamination on apples.

Chao et al. [28] conducted multi-spectral inspection based onfuzzy logic detection algorithms for differentiation of wholesomeand unwholesome chickens.

Yoon et al. [29] differentiated fecal and non-fecal poultry car-casses by HSI combined with kernel density estimation. Theyalso developed a prototype line-scan HSI system for inspectionof poultry carcasses with fecal material and ingesta, and pro-vided a commercially viable imaging platform for fecal detection[30].

Research on determining the total viable count of chilled porkwas carried out by Peng et al. [31]. The study showed that HSI isa valid tool for assessing the quality and safety properties of chilledpork during storage. Tao et al. [32] applied multi-linear regression(MLR) models for prediction of pork tenderness and Escherichia colicontamination based on hyperspectral scattering technique. Now-adays, the problem of pesticide residues has seriously influencedfood safety. Hu et al. [33] carried out an initial experiment to detect

pesticide residues on fruit surfaces by using laser HSI, and offeredan encouraging method for pesticide-residue detection.

HSI, as a powerful technique that combines information aboutspatial distribution and chemical composition, shows the poten-tial for assessing food-quality and safety properties. The HSI tech-nique has so far been widely applied in detection of defectivefruits, damaged mushrooms, meat products and teas. Amongthose, contamination detection of meat products seems to attractgreat attention. Image-processing methods or mathematic algo-rithms, such as ANNs and MLR, have been attempted in thosestudies, leading to good results. The hypercube contains bothspectra information and image information, which is a largeamount of data. Abundant and complex data processing con-sumes much time. At present, HSI seems more appropriate toserve as a research tool to develop multi-spectral imaging sys-tems based on only a few spectral bands for rapid on-line appli-cations [34]. Characteristic pictures obtained according to theselected wavelengths are generally enough to analyze results,and can markedly reduce the data-processing time. Future devel-opments in HSI equipment manufacture, such as lower cost pur-chases and improvements in processing speed, will enhance theadoption of this emerging platform for quality and safety controlin food industry [8].

Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274 265

2.2. Magnetic resonance imaging

Nuclear magnetic resonance (NMR) is a unique technologythat measures the magnetic properties of spins that can then berelated to the physical or chemical properties of subjects. NMRis the physical process that the nucleus, whose magnetic momentis not zero, resonantly absorbs radiation of a certain frequencyunder external magnetic field. Detectors detect and receive theNMR signals released as electromagnetic radiation; these signalscan then be sent to the computer and be converted into the im-age through data processing. MRI machines make use of the factthat food tissue contains lots of water which gets aligned in alarge magnetic field. Each water molecule has two hydrogen nu-clei or protons. When food is put in a powerful magnetic field, theaverage magnetic moment of many protons becomes aligned withthe direction of the field. A radio frequency transmitter is brieflyturned on, producing a varying electromagnetic field. This electro-magnetic field has just the right frequency, known as the reso-nance frequency, to be absorbed and flip the spin of theprotons in the magnetic field. After the electromagnetic field isturned off, the spins of the protons return to thermodynamicequilibrium and the bulk magnetization becomes re-aligned withthe static magnetic field. During this relaxation, a radio-frequencysignal is generated and can be measured with receiver coils.Information about the origin of the signal in 3D space can belearned by applying additional magnetic fields during the scan.A 3D image is compiled from multiple 2D images, which are pro-duced from any plane of view. The image can be rotated andmanipulated to be better able to detect tiny changes of structureswithin the food object. These fields, generated by passing electriccurrents through gradient coils, make the magnetic field strengthvary depending on the position within the magnet. Because thismakes the frequency of the released radio signal also depend onits origin in a predictable manner, the distribution of protons inthe food can be mathematically recovered from the signal, typi-cally using the inverse Fourier transform. In the images, each pix-el value reflects the NMR-signal intensity of a voxel in themeasured material, which relates with the resonance densityand the two main parameters (i.e., relaxation time: T1 and T2).MRI shows the image of the object structure making its physicaland chemical information visible.

In brief, the MRI system includes:

� the magnet and power-supply equipment that can produce awide range of uniform, stable and constant magnetic field;� a set of gradient magnetic field coil, a controller and power-dri-

ven equipment;� a radio-frequency (RF) system;� a computer system with large storage capacity for data collec-

tion and processing;� some auxiliary equipment.

Food displays a gigantic range of compositional and structuralcomplexity and heterogeneity, which explains its quality andsafety properties. Non-invasive techniques are then required ininvestigating such products to provide quantitative informationon the spatial organization [35]. Magnetic resonance, alreadyproved both technologically and commercially in clinical settings,was recently attempted in a laboratory test and introduced com-mercially into the on-line food-processing environment.

From NMR parameters, such as proton density and relaxationtime, many food characteristics can be quantified, including deter-mination of chemical composition, and quantification of the struc-ture [35]. Applications of NMR technique in food detection aremainly in the aspect of moisture, oil, carbohydrate, protein, andother quality parameters.

Qualitative and quantitative proton MRI techniques have beenapplied to food-quality assessment. Monitoring of fruit ripeningin persimmon [36], citrus [37] and oil palm fruit [38] has beenstudied, to detect tissue structure, water status of internal proper-ties or fungal inhibition using the MRI technique.

Cornillon et al. [39] studied the characterization of water mobil-ity and distribution in low moisture cereals and cookies, such ascorn flakes, chocolate-chip cookies, soft caramel candies and cornstarch/water systems, and they indicated that water mobilitychanged due to various chemical interactions in the system. Mac-Millan et al. [40] determined the oil and water contents in Frenchfries using MRI based upon the distribution of relaxation times.Thybo et al. [41] studied the use of MRI in determination of drymatter content in potatoes and identification of different varietiesof potato samples, in which the correlations between dry mattercontent and the MRI data were investigated using partial least-squares regression, but the results of the correlation coefficientwere not satisfactory. Recently, the use of MRI technique in deter-mination of fat content in meat was also conducted [42,43].

The MRI technique has proved to be a good tool for monitoringthe freezing process. Hills et al. [44] assessed the potential of theMRI technique to monitor the subcellular and intercellular redistri-bution of water non-invasively during drying and freezing ofparenchyma apple tissue. Using the changes in the distribution ofNMR water-proton transverse relaxation times, they found thatfreeze-drying apple tissue gave much lower water contents thanfluidized bed drying, but the NMR data confirmed that it destroyedmembrane integrity and caused cell-wall collapse.

Hindmarsh et al. [45] applied the MRI technique to visualize thefreezing process of sucrose solution. In a similar study, Mahdjoubet al. [46] used the MRI technique to monitor ice formation of a20% w/v sucrose solution during the freezing process, and to eval-uate quantitatively the glass-transition temperature and the dura-tion of crystallization phenomena. The results indicated that MRI,as a suitable tool, can image the phase behavior of sucrose solu-tions during cooling.

Other applications of MRI in food-quality inspection have alsobeen reported. Otero et al. [47] employed MRI to study the extentof the damage caused by relatively low pressures in strawberry.Infestation of apple fruits by the peach fruit moth was studiedusing MRI apparatus, and discrimination between sound and in-fested fruit was successful [48].

Compared with normal camera images, there are several advan-tages with NMR images, such as clear contrast, especially betweenfat and connective tissue, and 3D analysis of samples [43]. Due tothe principle of MRI, images are converted from electromagneticsignals that represent internal information of samples, so it hasgreat advantages in chemical-composition determination. As bodytissue contains a lot of water, MRI, which depends on protons (1Hnuclei), has great potential for food-quality assessment.

In conclusion, MRI has mainly been used for QC of various foodsinvolving cereal, meat product, fruit and vegetable, with the major-ity concerning the state of water that is correlated to food quality[42–48].

2.3. Soft X-ray imaging

X-ray, also called roentgen ray, is electromagnetic radiationwith the wavelength range 0.01–10 nm. The photon energy of anX-ray is in range 0.1–120 keV, which leads to strong penetrability.X-ray, similar to other electromagnetic waves, can show the fol-lowing phenomena: reflection, refraction, scattering, interference,diffraction, polarization and absorption. Usually, X-rays whosephoton energy is up to about 10 keV (10–0.10 nm wavelength)are classified as ‘‘soft’’ X-rays, and those of 10–120 keV (0.10–0.01 nm wavelength) are ‘‘hard’’ X-rays, due to their penetrating

Fig. 3. Principle of soft X-ray imaging (a) and the soft X-ray inspection system (b).

266 Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274

abilities. As hard X-rays pollute food, only the soft XRI technique isused in food inspection.

The principle of soft XRI inspection is based on the density ofthe product and the contaminant, as shown in Fig. 3(a). As an X-ray penetrates a food product, it loses some of its energy. A densearea, such as contaminant, will reduce the energy even further. Asthe X-ray exits the product, it reaches a sensor. The sensor thenconverts the energy signal into an image of the interior of the foodproduct. Foreign matter appears as a darker shade of grey thathelps to identify foreign contaminants. The soft X-ray inspectionsystem, as shown in Fig. 3(b), mainly comprises a computer-con-trolled X-ray generator (i.e. X-ray source tube), a line-scanningsensor for X-ray detection, conveying belt, stepping motor, im-age-acquisition card and computer. As a rapid, non-invasiveassessment technique, XRI also produces 3D information that canbe manipulated numerically. In contrast to MRI, soft XRI is rela-tively cheap to use and simple in accessibility and material restric-tions, such as ferromagnetic metals [49]. When investigating theinternal condition of foods, X-ray is most effective [50]. X-rayshave strong penetration ability, so the image can directly reflectinternal defects of food and agriculture products, and structuralorganization changes in quality. The XRI technique has great po-tential in detecting the internal quality of animal products, andhas been widely used in the food industry for the inspection of foodquality and safety.

Maintaining top quality is the key to success in the food busi-ness. XRI has shown potential for food-quality classification. Sha-hin et al. [51] discussed apple classification based on surfacebruises using the XRI technique combined with ANN. Kim et al.[52] detected internal water-core damage in apples using X-rayimagery with neural network classifier. Shahin et al. [53] also ap-plied XRI to detect internal defects in sweet onion.

X-ray, as a novel method has also been employed for the eval-uation of frozen products. Mousavi et al. [54] demonstrated thecapability of an X-ray micro CT system as a non-destructive tech-nique to characterize the ice-crystal microstructure of mycopro-tein products after freezing. They found that the dendrite spacingof ice crystals was related to the freezing conditions of the mate-rial, and that measurements of the voids by X-rays gave the sameresults as the measurements of conventional microscopy. Mousaviet al. [55] also investigated ice crystals formed during the freezingof a number of foods qualitatively and quantitatively using X-ray

micro-CT. The XRI could track the change in the material micro-structure as the freezing rate decreased away from the freezingsurface. It is possible to use this technique to determine ice-crystalparameters, for example, size area and width.

In fruit-storage control, Mendoza et al. [56] reported their studyon the multiscale structure of the pore-size distribution based onXRI technology. In this work, multifractal analysis (MFA) servedas an appropriate tool for characterizing the internal pore-size dis-tribution of apple tissue.

Applications of XRI in food manufacture were also reported.Kraggerud et al. [57] developed image-processing methods basedon X-ray instrument for the control of eye formation of cheesethroughout the ripening period.

Considering food-safety issues, foreign objects in food productsremain a danger to human health. In 2000, Tao et al. detected bonefragment in chicken fillets using thickness-compensated XRI [3]. Toavoid the influence of the uneven thickness of the meat fillet, Taoet al. [58] proposed an adaptive thresholding method for imagesegmentation in inspection of deboned poultry.

Kwon et al. [59] used XRI technique to discriminate foreign ob-jects of diverse types and sizes implanted in some packaged dryfoods.

Grain with fungal infection can be poisonous to consumers aswell as to animals, if used as feed. The potential of soft XRI to de-tect fungal infection in wheat was investigated by Narvankaret al. [60]. Healthy wheat kernels and kernels infected with thecommon storage fungi, namely Aspergillus niger, A. glaucus group,and Penicillium spp., were scanned using a soft XRI system andalgorithms were developed to extract the image features forclassification.

Fish bones are frequently ingested foreign bodies encounteredin foods. Mery et al. [61] developed an X-ray machine vision ap-proach to detect fish bones in fish fillets automatically; the meth-odology yielded a detection performance of 99%.

Compared with other imaging techniques, soft XRI technologyis based on sample density, and is superior in external contamina-tion detection and internal tissue distribution in food. Althoughwith a wide detection range, XRI cannot detect all kinds of foreignobjects. In general, foreign objects whose density is similar to thatof water cannot be easily recognized by the XRI technique. Forexample, such objects, including hair, paper and plastics, cannotbe detected using the XRI technique.

Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274 267

2.4. Ultrasound imaging

Ultrasound implies mechanical waves at frequencies above20 kHz, which is beyond the upper limit of the human auditoryacoustic frequency range (viz 20–20000 Hz). They are propagatedby vibration of the particles in the medium and may be reflectedand transmitted when they pass from one medium to another[62]. Detailed information about the different physical propertiesof materials can be acquired through the amount of energy re-flected or transmitted through materials depending on their rela-tive acoustic impedances. In addition, the time-of-flight andvelocity could also indicate a material property or changes in mate-rial characteristics, since ultrasound velocity depends on the den-sity and the elastic property of the medium [63]. Like lightwaves, incident ultrasound captures objects, and ultrasound en-ergy attenuation differs for the internal structure of an object toproduce a different echo, which leads to a series of points of lightdisplayed on the screen, that is, the ultrasound image. Primarily,the image contrast depends on differences in densities and speedsof sound, because these properties determine the scattering andthe reflectivity of tissue.

Since the 1960s, UI has undergone considerable developmentdue to the rapid development of modern electronic technology,computer technology and signal-processing techniques. There areseveral different modes of ultrasound, including A-mode (ampli-tude mode), B-mode (brightness mode), C-mode, Motion mode,Doppler mode, Pulse-inversion mode and Harmonic mode, inwhich a B-mode instrument has become the most commonly useddiagnostic ultrasound equipment. Compared with other imagingtechniques, UI is cheap, easy and without complicated post-im-age-processing procedures [62]. UI is a mature medical technology.It is said that more than a quarter of all clinical imaging proceduresused ultrasound in the UK and ultrasonic scans performed aremore popular than CT X-ray, MRI and radionuclide scanning [4].In recent years, UI proved its merit as one of the most promisingtechniques for food-quality and safety assessment due to its non-destructive nature, its rapidity, and its on-line potential.

The UI technique has gradually been applied to the detection ofagricultural and livestock products since 1990s. In food-qualityinspection, the microbial integrity of flexible food packages usuallydepends on a zero-defect level in the fused seam seal. Frazier et al.[64] presented two new processing techniques – radio-frequencysampling (RFS) technique and radio-frequency correlation (RFC) –that more reliably reveal smaller channels (approximate to 6 lmin diameter) than the previously developed pulse-echo backscat-tered amplitude integral (BAI) imaging technique. They showedthat RFS and RFC were superior to the BAI method for reliablydetecting channel defects less than 38 lm in diameter. Also, thecorrelation coefficient calculated over a short segment of the RFsignal performed computationally efficiently for both plastics andaluminum-foil trilaminate films with channels of widths 6 lm,10 lm, 15 lm, 38 lm, and 50 lm filled with water or air in termsof detection rates, image contrast, and contrast-to-noise ratio(CNR).

Shah et al. [65] developed a real-time approach for detectingseal defects in food packages using UI. They formed ultrasoundimages using a new raster scanning geometry, which simulatedcontinuous motion, based on the previously developed pulse-echoBAI mode imaging technique. The results showed that the rasterscanning geometry was feasible for on-line inspection. UI has beenapplied to fat-content detection of animal products for its discern-ment of soft tissue. Youssao et al. [66] studied the potential of UItechniques for predicting the lean content of carcasses. Correlationbetween ultrasound data about back-fat thickness, longissimusmuscle depth and longissimus area acquired from 335 pigs, andthe lean meat content estimated by a Capteur Gras/Maigre was

conducted. Although the result of prediction was not quite satisfac-tory, real-time ultrasound could be a tool used to predict the com-position of pig carcasses.

Also, works that focused in observing variations in consistencywithin starch-based liquids were conducted using non-contact UI.Harron et al. [67] predicted quality measures in beef cattle usingUI. Gan et al. [68] introduced a novel air-coupled ultrasonic inspec-tion system, in which ultrasonic signals over a reasonable band-width were generated in air using capacitance transducers withpolymer membranes. Various non-contact measurements for foodcontainers were demonstrated, including the inspection of varia-tions in consistency within starch-based liquids within a micro-waveable food container, and the inspection of liquid level anddetection of foreign objects in polymer-based soft-drink bottles.

Gan et al. [69] also detected physiochemical changes and den-sity variations in food using electrostatic transducers and signal-processing techniques in through-transmission mode.

In food safety, UI serves as a promising tool as well. When anultrasound wave travels through a non-uniform medium, part ofthe wave changes its initial direction and propagates separatelyfrom the original incident wave, distorting and interfering withthe original wave, in what is called scattering [62]. In the contextof foreign objects, such as bone, glass, or metal fragments in foodproducts, the amount of energy reflected and transmitted is ex-pected to be enormously different at the interfaces of the host tis-sue and the foreign object depending on their relative acousticimpedances. Previous studies [70,71] used contact ultrasoundmeasurement techniques with piezoelectric transducers to detectand to classify foreign bodies in commercial food samples. Choet al. [62] detected foreign objects and internal disorder in cheeseand poultry materials using a new non-contact air-instability com-pensation UI technique. Non-contact ultrasound images of bone-less chicken breast and cheese with a variety of foreign objectswere acquired and quantitative analysis of ultrasound parametershad to be performed. The results demonstrated the potential of thenon-contact UI technique for non-destructive, rapid detection offoreign objects. Pallav et al. [72] performed research to detect for-eign bodies and additives within food products using an air-cou-pled system that would allow use in manufacturing plants onproduction lines for non-contact sampling.

The principle of UI is similar to that of soft XRI, MRI, and HSI inthat they are all based on the amount of energy reflected and trans-mitted through materials, whereas the former is mechanical waveenergy and the latter are electromagnetic wave energy. The appli-cation of UI is mainly concentrated on food packaging and foreignobjects, and the results show better capability with the air-coupledsystem. Non-contact UI is emerging and tending to be a substitutein on-line detection of foreign bodies and additives.

2.5. Thermal imaging

TI is an emerging, non-invasive analytical tool suitable for thefood industry. The basic principle of TI is that all materials emitIR radiation, which is a band of invisible light found on the electro-magnetic spectrum with wavelengths of 0.75–100 lm. IR radiationcan be divided into five regions: near (0.75–2.5 lm), short wave(1.4–3 lm), mid (3–8 lm), long wave (>8 lm) and extreme (15–100 lm). TI is a non-invasive, non-contact system of recordingthermal distribution by measuring IR radiation emitted by a bodysurface to produce a pseudo image of the temperature distributionof the surface [73]. Among the IR radiation emitted by materials,radiation from the short-wave to long-wave IR (1.4–15 lm) canbe detected by TI systems. Typical long-wave IR systems showmaximum sensitivity around room temperature, while mid-waveIR systems show peak sensitivity at much higher temperatures(e.g., 400�C). Besides, NIR imaging systems also exist, but they

268 Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274

measure diffuse reflected radiation rather than emitted radiation.Thermography, which measures a large number of point tempera-tures of the target surface, is a powerful tool for visualizing andanalyzing targets with thermal gradients.

Temperature and emissivity are the two factors that determinethe amount of radiation emitted by an object. Emissivity is definedas the ratio of energy emitted from an object to that emitted from ablack body, varying from 0 (perfect white body) to 1 (perfect blackbody). The emissivity varies with the surface condition of the ob-ject and also with temperature and wavelength. The surface tem-perature of the object is estimated based on the IR energyemitted by it, but cannot be determined by the total radiant emit-tance alone. The relationships between temperature, emissivityand energy emitted by the object were described by Gowen et al.[74]. According to the black body radiation law, IR radiation isemitted by all objects above absolute zero. The TI technique en-ables detection of bodies whose surface temperature is distin-guishable from their backgrounds with or without visibleillumination.

Thermal images can be obtained using passive or active TIsystems. Passive thermography refers to TI without applyingany external energy to the object; the features of interest arenaturally at a higher or lower temperature than the background.Passive thermography has many applications, such as surveil-lance of people on a scene and medical diagnosis (specificallythermology), and is useful for non-contact temperature measure-ment of foods during processing. Active thermography requiresthe application of thermal energy to produce a thermal contrastbetween the feature of interest and the background. This type ofthermography is necessary in many cases, given that the in-spected parts are usually in equilibrium with the surroundings,and can be used to detect surface and sub-surface defects infoods.

TI systems typically comprise the following components: cam-era, an optical system (e.g., focusing lens, collimating lenses, andfilters), detector array (e.g., microbolometers), signal processing,and image-processing system. TI does not require an illuminationsource, but integrated systems for active thermography measure-ments contain a heating or cooling unit to provide a thermal differ-ential. Specialized TI cameras use focal plane arrays (FPAs) thatrespond to longer wavelengths (mid- and long-wavelength IR).The newest technologies use low-cost, uncooled microbolometersas FPA sensors. In TI cameras, the IR energy emitted from an objectunder investigation is converted into an electrical signal via IRdetectors and displayed as a monochrome or color thermal image.The image-acquisition speed of the approach may be high enough(e.g., 50–60 images/s) to explore rapidly changing thermal condi-tions [75].

All objects above the absolute zero temperature (0 K) emit IRradiation. TI was originally developed for military applicationsand for surveillance in night vision. With technological advances,especially in computer analytical tools with digitalized high-reso-lution imaging, TI was increasingly used in various fields, includingmedicine, materials science and fire safety. TI systems are also suit-able for the food industry, due to their portability, real-time imag-ing, and non-contact temperature-measurement capability.Increasing demands for objectivity, consistency and efficiencywithin the food industry have necessitated the introduction ofcomputer-based image-processing techniques. Thermal images,or thermograms, where both spatial and temporal temperaturedistribution patterns are visually displayed from the body underinvestigation have potential applications for food-product QA,safety profiling and authenticity compliance. Recent advancesand potential applications of TI for food-quality and safety assess-ment, such as temperature, bruise validation, foreign-body detec-tion and grain-quality evaluation are reviewed.

Traditionally, temperature control and monitoring in a food-manufacturing process is critical to ensure product quality duringproduction, transportation, storage and sales. The amount of radi-ation emitted by an object increases with temperature, so ther-mography visualizes variations in temperature. Goedeken et al.[76] designed a continuous temperature-measurement systemfor food materials in a microwave oven during heating by directTI. Ibarra et al. [77] estimated the internal temperature of chickenmeat immediately following cooking using a TI technique based ona multilayer neural network; and they validated the potential of TIfor monitoring the doneness of chicken meat on a conveyor belt.Varith et al. [78] detected bruises in apples by TI by discriminatingsurface temperature between bruised and sound tissues, and theexperiment indicated that the temperature differences betweenbruised and sound tissues were due to the differences in thermaldiffusivity, not emissivity differences. Food-quality inspection byTI have been reported in many other areas, such as tomatobruise-damage (soft spots) detection [79], ice nucleation and freez-ing of cauliflower plant [80], meat quality and slaughter-line defectdetection [81,82], walnut process monitoring [83], citrus dryingprocess [84] and grain-quality detection [85,86].

The presence of foreign bodies in food is a major food-safety is-sue. TI is presented as a new method for inspecting food samples,using thermographic images to distinguish foreign bodies and foodmaterials due to differences in their thermal properties. Experi-ments to obtain well-contrasted thermographic images of differentfood and foreign bodies were conducted. Meinlschmidt et al. [87]gave an introduction to the possibility of detecting foreign bodiesin food by IR thermography with specially adapted algorithmsusing statistical or morphological analysis. Meinlschmidt et al.[2] carried out more detailed investigation to develop suitablealgorithms for automatic detection of foreign bodies after the ini-tial promising tests, and numerous experiments in image process-ing and pattern recognition were exploited. Ginesu et al. [88]conducted experiments on foreign-body detection using activethermography, and the results were promising. As all objects aboveabsolute zero emit thermal IR energy, TI cameras convert the en-ergy in the IR wavelength into a visible light display.

Thermography shows a visual picture, so temperatures over alarge area can be compared, and it is capable of catching movingtargets in real time. As a non-destructive, non-contact test method,it can be used to measure or observe in areas inaccessible or tohazardous for other methods. Thermography has limitations anddisadvantages:

� first is the price of TI camera is higher than visible-spectrumcounterparts;� second, the image is difficult to interpret accurately when based

upon certain objects, specifically objects with erratictemperatures;� most cameras have ±2% accuracy or worse in measurement of

temperature and are not as accurate as contact methods;� thermal interference from the ambient environment can also

influence the test results.

With improvements in technology, the TI technique is becom-ing more reliable, user friendly, accurate and cost effective. Giventhe potential applications discussed, increased adoption of thistechnology by food industry is likely.

2.6. Fluorescence imaging

Fluorescence is the emission of light by a substance that has ab-sorbed light or other electromagnetic radiation. It is a form of lumi-nescence. In most cases, the emitted light has a longer wavelength,and therefore lower energy, than the absorbed radiation.

Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274 269

Fluorescence will disappear immediately without the incidentlight. There are two kinds of fluorescence, which are auto-fluorescence and fluorescence with the help of fluorescentpigment. The most striking examples of fluorescence occur whenthe absorbed radiation is in the ultraviolet region of the spectrum,which is invisible to the human eye, but the emitted light is in thevisible region. The process of FI is as follows:

� objects produce fluorescence when absorbed incident light;� an imaging equipment (e.g., electron-multiplying charge-cou-

pled-device) catches the light emitted and produces images;� image-processing methods are then implemented.

Commonly used for FI, laser confocal scanning microscopy(LCSM), integrating visualization and microscope, is suitable formicrobial detection in food samples [89]. Besides, hyperspectralor multi-spectral FI also shows potential for food-quality andsafety assessment.

The optical sensing technique is based on fluorescence that isgenerally regarded as more sensitive for optical sensing toolscompared with reflectance techniques. Some researchersconducted the visualization of the distributions of gluten, starchand air bubbles in dough by fluorescence fingerprint imaging,and it proved to be a powerful visualization tool [90,91]. Also,Adedeji et al. [92] evaluated deep-fat fried chicken-nugget battercoating using LCSM, which obtained images in fluorescence modeto show fat distribution. Image analysis showed how fatdistribution was affected by frying time, temperature and productdepth.

FI techniques demonstrated that even minute animal feces, notclearly visible to human eye, could be detected. Byoung-Kwan et al.[93] developed a laser-induced FI system (LIFIS) for detectingfeces-contaminated poultry carcasses in ambient light. Yang et al.[94] evaluated three multi-spectral algorithms, separation meth-ods using a boundary based on the manual drawing of a simplestraight-line (OA), the shape of the cluster of apple pixels (OL),and the shape of the cluster of feces pixels (OF) for identificationof fecal contamination on Golden Delicious apples.

Fig. 4. The odor-imaging tool used in d

Jun et al. [95] carried out a study determining a minimal num-ber of spectral fluorescence bands suitable for detecting microbialbiofilms on food-contact surfaces. The result showed that a single-band image at 559 nm was able to detect the biofilm spots onstainless steel. Qin et al. [96] investigated the potential of a light-emitting-diode-induced FI (LED-induced FI) technique for rapidinspection of organic residues on equipment surfaces in poultry-processing plants.

In summary, the FI technique, under certain conditions, has thepotential for usage for detection in food quality and safety, butconventional FI has certain limitations in analysis of food qualityand safety. Unlike transmitted and reflected light imaging tech-niques, FI only allows observation of the specific structures thathave been labeled for fluorescence. Accordingly, it cannot be suffi-cient for measuring some quality attributes (e.g., fruit titratableacidity, sugar or soluble solid content), since not all materials canbe excited to fluoresce; moreover, scattering of the incident lightaffects the fluorescence signal. In future, we can integrate the FItechnique with the other imaging tools, such as microscopy imag-ing and HSI, to broaden its application in evaluation of food qualityand safety.

2.7. Odor imaging

OI based on colorimetric sensor array, is the latest new technol-ogy for non-visible matter (odor) detection. The basic principle ofthe OI technique is to utilize the color change induced by reactionbetween volatile material and an array of chemically-responsivedyes upon ligand binding for chemical-vapor detection and differ-entiation. Chemically-responsive dyes were selected according totheir sensitivity to volatile compounds that need to be detected.Metalloporphyrins are usually a natural choice for the detectionof metal-ligating vapors as they have open coordination sites foraxial ligation and large spectral shifts upon ligand binding, andespecially intense coloration [97]. Common pH indicator dyes havebeen selected for their color change in response to changes in theproton acidity or alkalinity of their environment [98]. Reversed-phase silica has been chosen as a non-interacting dispersion

etection of food quality and safety.

270 Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274

medium for the chemically-responsive dye array, and as a suitablesurface for diffuse reactance spectral measurements. The colori-metric sensor array can be made by printing selected dyes on a re-versed-phase silica-gel plate. The array responses originate fromselective and specific interactions between the vapor of interestand the metalloporphyrin library. A color change profile for eachobject can be obtained by differentiating the images of the sensorarray before and after exposure to the volatile compounds of ob-jects. The digital data representing the color change profiles arethen analyzed using data-processing methods and pattern-recog-nition methods [99]. Fig. 4 shows the functional prototype of theOI system. First, the sensors array is captured by a charge-coupleddevice (CCD) camera before exposure to the sample, and an initialimage of the sensor array is achieved. Next, the array is exposed tothe sample, and starts to change color in response to volatile or-ganic compounds (VOCs) released from the sample. On reachingnearly complete equilibration, the sensor array is captured againby CCD camera, and gets a ‘‘final’’ image. We can get a color-differ-ence image by subtracting the ‘‘initial’’ image from the ‘‘final’’ im-age, and the difference image provides a color-change profile thatis a fingerprint characteristic of the VOCs in sample. The colorimet-ric sensor array in this E-nose system can generate a characteristicfingerprint to an odor stimulus. Patterns or fingerprints fromknown odors are often employed to construct a database and traina pattern-recognition system so that unknown odors can subse-quently be classified and identified. Hence, this imaging tool usedin evaluation of food quality and safety usually needs the help ofmultivariate calibration [e.g., principal component analysis (PCA),linear discriminant analysis (LDA), hierarchical clustering analysis(HCA), or artificial neural network (ANN)].

The conventional artificial olfactory system to detect and to dif-ferentiate between chemically diverse analytes is concentrated onnon-specific chemical interactions to detect non-coordinating or-ganic vapors. The colorimetric sensor array permits the visual iden-tification of a wide range of ligating (amines, alcohols, ethers,phosphites, phosphines, thioethers and thiols) and even weaklyligating (arenes, halocarbons and ketones) vapors [97]. Takingadvantage of the large color changes induced in metalloporphyrinsupon ligand binding, we are able to obtain unique color-changesignatures for analytes; and, again, with the hydrophobic natureof reversed-phase silica, water vapor does not affect the perfor-mance of the device. Besides, this easy colorimetric technique withhigh accuracy minimizes the need for extensive signal-transduc-tion hardware.

The OI technique based on a low-cost, sensitive colorimetricsensor array for detection and identification of VOCs has foundapplication in food-quality and safety assessment. Evaluation oftea quality and classification of tea variety were attempted bythe OI technique in both liquid and gas phases [100]. Zhang et al.[101] demonstrated the potential of colorimetric sensor array tech-nology for QC applications of soft drinks. 25 chemically-responsivedyes were selected for a colorimetric sensor array, and statisticaland chemometric methods, including PCA and HCA, were em-ployed for data analysis. Chan et al. [102] discriminated differentbrands of a bottled water by use of a colorimetric sensor array.

The excellent potential of colorimetric sensor array for complexsystems analysis was demonstrated by Suslick et al. [103]. They de-tected and identified coffee aromas based on PCA and HCA.

Studies on alcohol analysis by colorimetric sensor array havebeen reported. Zhang et al. [104] analyzed beers in both liquidand gas phases using a colorimetric sensor array; PCA and HCAwere used for data analysis and performed well. Recently, the OItechnique combined with PCA and HCA was attempted in classifi-cation of Chinese liquor types or geographic origins [105–108].

Also, evaluation of meat freshness was an application of an OItechnique. Huang et al. [98] developed an olfaction system based

on a colorimetric sensor array for fish-freshness evaluation. Salinaset al. [109] attempted to monitor chicken-meat freshness by meansof a colorimetric sensor array. Lim et al. [110] employed a colori-metric sensor array for identification and quantification of sugarsand related compounds. Chen et al. [111] attempted a portableOI system (namely, a colorimetric sensor array) to monitor vinegaracetic fermentation. The colorimetric sensor array was fabricatedby printing 15 chemically-responsive dyes (i.e. nine porphyrins/metalloporphyrins and six pH indicators) on a C2 reverse silica-gel flat plate, and it was successfully used to monitor vinegar aceticfermentation with the help of multivariate calibration.

Differing from other imaging techniques, the OI technique dis-plays signal according to the aroma (volatile substance) of materi-als, so only odorous food are suitable for this inspection tool. Also,the OI tool is a newly developing technology with some problemsto be solved (e.g., the selection of responsive dyes for colorimetricsensor array). Despite that, the OI technique is sensitive, conve-nient and with easy operation, when used for inspection of foodquality and safety.

3. Technical challenges and trends

In order to observe some trends on how the imaging techniquesare applied and briefly review all the references, we summarize theon-line or off-line solutions and food quality or food safety aspectsreviewed in this paper in Table 1. The references are groupedaccording to imaging mode. For each reference, Table 1 lists thefood product and the main properties that are inspected. The majorapplication area of the imaging techniques, as used in the refer-ence, is marked in the column corresponding to the applicationarea. With the columns ‘‘Food quality’’, ‘‘Food safety’’, ‘‘Off-line’’and ‘‘On-line’’, we indicate the location of the reference on thescale. Based on the reviewed references and observed trends, wesummarize the recent developments and point toward some po-tential future directions in the field. Most of the literature in thisreview comprises papers published since 2005. As shown in Ta-ble 1, most of the emerging imaging techniques have in recentyears become valuable tools to inspection of food quality andsafety. In particular, with the development of digital imaging tech-nique, computer hardware processing power and image-process-ing algorithm, together with the application of specific researchdevelopments described in this review, some emerging imagingtechniques have enabled the on-line and off-line inspection of foodproducts with respect to food quality and safety. Furthermore,most of the imaging techniques have become valuable throughoutthe entire value chain of food industries. The general trend seemsto point toward the use of imaging techniques to optimize produc-tion cost-effectiveness and product quality in many aspects. Basedon the referenced works and the observed trends from Table 1, wedescribe the technical challenges and future outlook for theseemerging imaging techniques to analysis of food quality and safety.

First of all, most of imaging techniques, apart from the OI tech-nique, have the capabilities for on-line application for food indus-try. Automation will continue to increase by being applied tomore applications that are not currently automated. As more pro-cesses become automated, further integration is possible betweenautomated processing units. This is a transition from automationof a single processing unit to a more coherent, potentially moreflexible automation of the entire processing chain in food indus-tries. In addition, when talking about how these advanced imag-ing technologies facilitate the automation of food applicationsand processes, the choice of suitable end-effectors, as system ele-ments in contact with the processed material, is also of greatimportance for the success of these emerging imaging techniquesused for automated inspection, handling, and processing of food.

Table 1Summary of imaging mode, product type, property studied, application area and references since 2000

Imaging mode Product Studied property Foodsafety

Foodquality

Off-line

On-line

Ref.

Hyperspectralimaging

Apple Detection of defects on applesp p

[10]Mushroom Sorting damaged mushroom

p p[11]

Kiwifruits Inspecting bruises on Kiwi fruitsp p

[12]Cooked turkey hams Classification of cooked turkey hams with different ingredients and

processing parameters

p p[13]

Avocados Assessing avocado qualityp p

[14]Tea Assessing tea quality

p p[15–20]

apple Sorting contaminated applesp p

[27]Chicken Differentiation of unwholesome chickens

p p[28]

Poultry carcasses Classification of fecal and non-fecal poultry carcassesp p

[30]Pork Determining the total viable count of chilled pork, pork tenderness and

E. coli contamination

p p[32]

Magneticresonanceimaging

Persimmon, citrus, oilpalm fruit

Monitoring of fruit ripeningp p

[36]

Cereals and cookies Study of water mobility and distributionp p

[39]Cooked white saltednoodles

Evaluating water status of cooked white salted noodlesp p

[112]

French fries Determining the oil and water contents in French friesp p

[40]Potato Determination of dry matter content in potatoes and identification of

different varieties of potato samples

p p[41]

Meat Determination of fat content in meatp p

[42,43]Sucrose solution Study of the freezing process of sucrose solution

p p[46]

Strawberry Study the extent of the damage in strawberryp p

[47]Apple Sorting sound apples and infested apples

p p[48]

X-ray imaging Apple Sorting apples with bruisesp p

[51]Apple Detecting internal water core damage in apples

p p[52]

Sweet onion Detecting internal defects in sweet onionp p

[53]Frozen products Investigating ice crystals

p p[54,55]

Fruit Study on multiscale structure of the pore-size distribution in fruitp p

[56]Cheese Control of eye formation of cheese

p p[57]

Chicken Detecting bone fragment in chicken filletsp p

[58]Packaged dry foods Discriminating foreign objects in some packaged dry foods

p p[59]

Wheat Detecting fungal infection in wheatp p

[60]Fish Detecting fish bones in fish fillets

p p[61]

Ultrasoundimaging

Food packages Seal defect detection in food packagesp p

[68]Pig Prediction of lean content of carcasses

p p[66]

Milk-based products Detecting physiochemical changes and density variations in foodp p

[69]Cheese, poultry Detecting foreign object and internal disorder in cheese and poultry

p p[62]

Cheese; chocolate;dough

Detecting foreign bodies and additivesp p

[72]

Thermal imaging Chicken Monitoring of doneness of chickenp p

[77]Apple Sorting bruised and sound tissues of apples

p p[78]

Tomato Detecting tomato bruise damagep p

[79]Meat Detecting meat quality

p p[81,82]

Walnut Monitoring walnut processp p

[83]Citrus Monitoring citrus drying process

p p[84]

Milk-based products;oil

Detecting foreign bodies in food; physiochemical changes and densityvariations in food

p p[87]

Fluorescenceimaging

Dough Study the distributions of gluten, starch and air bubbles in doughp p

[90,91]Chicken Study the fat distribution in chicken

p p[92]

Poultry Detecting feces-contaminated poultry carcassesp p

[93]Apple Identifying fecal contamination on apples

p p[94]

Food contact surfaces Detecting biofilm spots on stainless steelp p

[95]Food equipmentsurfaces

Inspection of organic residues on processing equipment surfacesp p

[96]

Odor imaging Tea Evaluating tea quality and classification of tea varietiesp p

[100,113]Soft drinks Quality control of soft drinks

p p[101]

Bottled water Discriminating different brands of bottled waterp p

[102]Coffee Identifying coffee aromas

p p[103]

Beer Classification of beersp p

[104]Chinese liquor Identifying Chinese liquor types or geographic origins

p p[105–108]

Meat Evaluating fish and chicken meat freshnessp p

[98,109]Vinegar Monitoring vinegar acetic fermentation

p p[111]

Sugar Identification and quantification of sugars and related compoundsp p

[110]

Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274 271

Flexible automation is seen as a way to address large naturalbiological variation of agricultural product/food specific applica-tions. Recently, there has been an ever-increasing demand fromspecific food processors for automated integrated systemsinvolving intelligent sensing (the imaging technique in certaincondition) and end-effectors. Such a flexible automation,

involving several processing operations, naturally implies achallenge in creating reconfigurable and adaptive image-acquisi-tion and image-processing systems – significantly morechallenging than detecting specific properties in a single specificspecies as presented to the imaging system in a single processingstage.

272 Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274

Second, food quality and safety contain various aspects involv-ing external factors (e.g., size, shape, color, gloss and consistency)and internal factors (e.g., texture, flavor, chemical, physical andmicrobial). Most of emerging imaging techniques, with use in spe-cific applications, cannot simultaneously solve all aspects of foodquality and safety. Thus, cost-effective solutions to challengingproblems of the future will make use of integration of multipleimaging technologies, imaging modes and even non-imaging sen-sors. For example, HSI is an emerging tool for food quality andsafety analysis; the spatial feature of HSI enables characterizationof complex heterogeneous samples, while the spectral feature al-lows identification of a wide range of multi-constituent surfaceand sub-surface features. However, the conventional HSI techniquehas limitation in analysis of microorganisms of concern in foodquality. With the help of the advantages from fluorescence micro-scopic imaging (FMI) and Raman microscopic imaging (RMI) inanalysis of food bacteria, the technical integration with HSI andFMI (or RMI) can add the capabilities in inspection of food bacterialcontamination to those food properties that can be inspected bythe conventional HSI. We anticipate that future technologicaldevelopment in the HSI system for food-safety analysis will pro-mote some novel imaging techniques, such as hyperspectral FMIor hyperspectral RMI. In addition, we can integrate the emergingimaging tools with the non-imaging tools, such as electronic noseor electronic tongue, to evaluate food quality and safety broadly.

As the objective assessment of food products, with respect tovarious properties, becomes available, there will be use for thisassessment in order to increase product homogeneity and enablegreater market and product differentiation. In particular, MRI hasthe potential to enable a complete understanding and utilizationof products – providing detailed information to automated inspec-tion and processing units, and also enabling on-line process under-standing and optimization of the entire food-process chain.However, the great limitations of this technology today lie in theexpense of the equipment and the challenges of image processing.

Finally, the major barrier to application of these emerging imag-ing techniques in the food industry is budget constraint. An image-processing system is still unviable in many potential applicationsbecause the cost is unacceptable. For example, a current MRI tool,only a laboratory set-up, has the potential to enable a completeunderstanding and utilization of products – providing detailedinformation to automated inspection and processing units, andalso enabling on-line process understanding and optimization ofthe entire food-process chain. To satisfy the need for cost-effective-ness, developing a cheap specific imaging system is especiallyimportant to food-quality and safety analysis. Given that the massof data obtained by these emerging imaging tools is complex andirrelevant for inspection of food properties, the next challenge isinterpretation and use of mass data. Image-processing speed is stilla bottleneck in heavy-duty real-time applications, failing to handlethe large data streams. Selecting an efficient, practical machine-learning algorithm is therefore a precondition for successful appli-cation of these emerging imaging tools in detection of food qualityand safety. On the one hand, developing adequate, efficient andaccurate image-processing algorithms can accelerate processingspeed to meet modern manufacturing requirements. On the otherhand, integrating image-processing algorithms into specializedhardware can significantly reduce test time. In brief, developinghigh-performance, low-cost imaging equipment will be a trendto its future practical usages in the food industries.

4. Conclusions

This review summarized emerging imaging technologies andtheir major applications to food-quality and safety inspection. This

review looked at the observed trends in their practical usages inthe food industries. On the basis of the observed trends, we alsopresented the technical challenges and future outlook for theseemerging imaging techniques. Given these attributes and the po-tential applications discussed, increased adoption of these emerg-ing techniques by the food industry for improved efficiency islikely.

Acknowledgements

This work has been financially supported by the National Natu-ral Science Foundation of China (31271875). We are also grateful tomany of our colleagues for stimulating discussion in this field.

References

[1] X. Yang, J. Fu, Z. Lou, L. Wang, G. Li, W. Freeman, Tea classification based onartificial olfaction using bionic olfactory neural network, Lect. Notes Comput.Sci. 3972 (2006) 343–348.

[2] P. Meinlschmidt, V. Maergner, Thermographic techniques and adaptedalgorithms for automatic detection of foreign bodies in food. Proc. SPIE,Thermosense XXV, 5073 (2003) 168–176.

[3] Y. Tao, J. Ibarra, Thickness-compensated X-ray imaging detection of bonefragments in deboned poultry-model analysis, Trans. ASAE 43 (2000) 453–459.

[4] H.D. Liang, J.A. Noble, P.N.T. Wells, Recent advances in biomedical ultrasonicimaging techniques, Interface Focus 1 (2011) 475–476.

[5] F.J. Señorans, E. Ibáñez, A. Cifuentes, New trends in food processing, Crit. Rev.Food Sci. Nutr. 43 (2003) 507–526.

[6] C.J. Du, D.W. Sun, Recent developments in the applications of imageprocessing techniques for food quality evaluation, Trends Food Sci. Technol.15 (2004) 230–249.

[7] J.R. Mathiassen, E. Misimi, M. Bondø, E. Veliyulin, S.O. Østvik, Trends inapplication of imaging technologies to inspection of fish and fish products,Trends Food Sci. Technol. 22 (2011) 257–275.

[8] A.A. Gowen, C.P. O’Donnell, P.J. Cullen, G. Downey, J.M. Frias, Hyperspectralimaging – an emerging process analytical tool for food quality and safetycontrol, Trends Food Sci. Technol. 18 (2007) 590–598.

[9] R. Díaz, L. Cervera, S. Fenollosa, C. Ávila, J. Belenguer, Hyperspectral system forthe detection of foreign bodies in meat products, Procedia Eng. 25 (2011)313–316.

[10] P. Mehl, K. Chao, M. Kim, Y. Chen, Detection of defects on selected applecultivars using hyperspectral and multispectral image analysis, Appl. Eng.Agric. 18 (2002) 219–226.

[11] A. Gowen, C. O’Donnell, M. Taghizadeh, P. Cullen, J. Frias, G. Downey,Hyperspectral imaging combined with principal component analysis forbruise damage detection on white mushrooms (Agaricus bisporus), J.Chemom. 22 (2008) 259–267.

[12] Q. Lu, M. TaNg, J. Cai, J. Zhao, S. ViTTayapaduNg, Vis/NIR hyperspectralimaging for detection of hidden bruises on kiwifruits, Czech. J. Food Sci. 29(2011) 595–602.

[13] G. ElMasry, A. Iqbal, D.-W. Sun, P. Allen, P. Ward, Quality classification ofcooked, sliced turkey hams using NIR hyperspectral imaging system, J. FoodEng. 103 (2011) 333–344.

[14] N. Maftoonazad, Y. Karimi, H. Ramaswamy, S. Prasher, Artificial neuralnetwork modeling of hyperspectral radiometric data for quality changesassociated with avocados during storage, J. Food Process. Preserv. 35 (2011)432–446.

[15] F. Jiang, X. Qiao, H. Zheng, Q. Yang, Grade discrimination of machine-friedLongjing tea based on hyperspectral technology, Trans. CSAE 27 (2011) 343–348.

[16] X.L. Li, Y. He, Classification of tea grades by multi-spectral images andcombined features, Trans. CSAM 40 (2009) 113–118.

[17] J.W. Zhao, Q.S. Chen, J.R. Cai, Q. Ouyang, Automated tea quality classificationby hyperspectral imaging, Appl. Opt. 48 (2009) 3557–3564.

[18] D. Wu, X.J. Chen, Y. He, Application of multispectral image texture todiscriminating tea categories based on DCT and LS-SVM, Spectrosc. Spect.Anal. 29 (2009) 1382–1385.

[19] X.L. Li, Y. He, Z.J. Qiu, Textural feature extraction and optimization in waveletsub-bands for discrimination of green tea brands, Proc. 17th Int. Conf. Mach.Learn. Cybern. 3 (2008) 1461–1466.

[20] X.J. Chen, D. Wu, Y. He, Study on discrimination of tea based on color ofmultispectral image, Spectrosc. Spect. Anal. 28 (2008) 2527–2530.

[21] J. Zhao, Q. Ouyang, Q. Chen, J. Wang, Detection of bruise on pear byhyperspectral imaging sensor with different classification algorithms, Sens.Lett. 8 (2010) 570–576.

[22] J. Gómez-Sanchis, E. Moltó, G. Camps-Valls, L. Gómez-Chova, N. Aleixos, J.Blasco, Automatic correction of the effects of the light source on sphericalobjects. An application to the analysis of hyperspectral images of citrus fruits,J. Food Eng. 85 (2008) 191–200.

Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274 273

[23] J.W. Qin, K.L. Chao, M.S. Kim, Evaluating carotenoid changes in tomatoesduring postharvest ripening using Raman chemical imaging, Sens. Agric. FoodQual. Saf. III 8027 (2011) 03–11.

[24] Q. Jun, M. Ngadi, N. Wang, A. Gunenc, M. Monroy, C. Gariepy, S. Prasher, Porkquality classification using a hyperspectral imaging system and neuralnetwork, Int. J. Food Eng. 3 (2007) 1–12.

[25] D. Barbin, G. Elmasry, D.-W. Sun, P. Allen, Near-infrared hyperspectralimaging for grading and classification of pork, Meat Sci. 90 (2012) 259–268.

[26] M. Kim, A. Lefcourt, K. Chao, Y. Chen, I. Kim, D. Chan, Multispectral detectionof fecal contamination on apples based on hyperspectral imagery: part I.Application of visible and near-infrared reflectance imaging, Trans. ASAE 45(2002) 2027–2038.

[27] C.C. Yang, M.S. Kim, K. Chao, S. Kang, A.M. Lefcourt, in: M.S. Kim, S.I. Tu, K.Chao (Eds.), Fast and Accurate Image Recognition Algorithms for FreshProduce Food Safety Sensing, Proc. SPIE, Orlando, FL, 2011, p. 8027.

[28] K. Chao, C. Yang, Y. Chen, M. Kim, D. Chan, Hyperspectral-multispectral line-scan imaging system for automated poultry carcass inspection applicationsfor food safety, Poult. Sci. 86 (2007) 2450–2460.

[29] S. Yoon, K. Lawrence, B. Park, W. Windham, Optimization of fecal detectionusing hyperspectral imaging and kernel density estimation, Trans. ASABE 50(2007) 1063–1071.

[30] S.C. Yoon, B. Park, K.C. Lawrence, W.R. Windham, G.W. Heitschmidt, Line-scanhyperspectral imaging system for real-time inspection of poultry carcasseswith fecal material and ingesta, Comput. Electron. Agric. 79 (2011) 159–168.

[31] Y. Peng, F. Tao, Y. Li, W. Wang, J. Chen, J. Wu, S. Dhakal, Rapid detection oftotal viable count of chilled pork using hyperspectral scattering technique,Int. Soc. Opt. Photonics 7676 (2010) 76760K1–76760K8.

[32] F. Tao, Y. Peng, Y. Li, K. Chao, S. Dhakal, Simultaneous determination oftenderness and Escherichia coli contamination of pork using hyperspectralscattering technique, Meat Sci. 90 (2012) 851–857.

[33] S. Hu, M. Liu, H. Liu, A study on detecting pesticide residuals on fruit surfaceby using laser imaging, Acta Agri. Univ. Jiangxi 6 (2006) 013.

[34] M. Kim, K. Chao, D. Chan, W. Jun, A. Lefcourt, S. Delwiche, S. Kang, K. Lee, Line-scan hyperspectral imaging platform for agro-food safety and qualityevaluation: system enhancement and characterization, Trans. ASABE 54(2011) 703–711.

[35] F. Mariette, Relaxation RMN et IRM: un couplage indispensable pour l’étudedes produits alimentaires, C.R. Chim. 7 (2004) 221–232.

[36] C.J. Clark, J.S. MacFall, Quantitative magnetic resonance imaging of ‘Fuyu’persimmon fruit during development and ripening, Magn. Reson. Imaging 21(2003) 679–685.

[37] G. Galed, M.E. Fernández-Valle, A. Martı́nez, A. Heras, Application of MRI tomonitor the process of ripening and decay in citrus treated with chitosansolutions, Magn. Reson. Imaging 22 (2004) 127–137.

[38] S.M. Shaarani, A. Cardenas-Blanco, M. Amin, N.G. Soon, L.D. Hall, Monitoringdevelopment and ripeness of oil palm fruit (Elaeis guineensis) by MRI and bulkNMR, Int. J. Agric. Biol. 12 (2010) 101–105.

[39] P. Cornillon, L.C. Salim, Characterization of water mobility and distribution inlow- and intermediate-moisture food systems, Magn. Reson. Imaging 18(2000) 335–341.

[40] B. MacMillan, H. Hickey, B. Newling, M. Ramesh, B. Balcom, Magneticresonance measurements of French fries to determine spatially resolved oiland water content, Food Res. Int. 41 (2008) 676–681.

[41] A.K. Thybo, H.J. Andersen, A.H. Karlsson, S. Dønstrup, H. Stødkilde-Jørgensen,Low-field NMR relaxation and NMR-imaging as tools in differentiationbetween potato sample and determination of dry matter content inpotatoes, LWT Food Sci. Technol. 36 (2003) 315–322.

[42] J.P. Wold, M. O’Farrell, M. Høy, J. Tschudi, On-line determination and controlof fat content in batches of beef trimmings by NIR imaging spectroscopy,Meat Sci. 89 (2011) 317–324.

[43] L. Ballerini, A. Hogberg, G. Borgefors, A.C. Bylund, A. Lindgard, K. Lundstrom,O. Rakotonirainy, B. Soussi, A segmentation technique to determine fatcontent in NMR images of beef meat, IEEE Trans. Nucl. Sci. 49 (2002) 195–199.

[44] B.P. Hills, B. Remigereau, NMR studies of changes in subcellular watercompartmentation in parenchyma apple tissue during drying and freezing,Int. J. Food Sci. Technol. 32 (1997) 51–61.

[45] J.P. Hindmarsh, C. Buckley, A.B. Russell, X.D. Chen, L.F. Gladden, D.I. Wilson,M.L. Johns, Imaging droplet freezing using MRI, Chem. Eng. Sci. 59 (2004)2113–2122.

[46] R. Mahdjoub, P. Chouvenc, M.J. Seurin, J. Andrieu, A. Briguet, Sucrose solutionfreezing studied by magnetic resonance imaging, Carbohydr. Res. 341 (2006)492–498.

[47] L. Otero, G. Préstamo, Effects of pressure processing on strawberry studied bynuclear magnetic resonance, Innovative Food Sci. Emerg. Technol. 10 (2009)434–440.

[48] T. Haishi, H. Koizumi, T. Arai, M. Koizumi, H. Kano, Rapid detection ofinfestation of apple fruits by the peach fruit moth, Carposina sasakiiMatsumura, larvae using a 0.2-T dedicated magnetic resonance imagingapparatus, Appl. Magn. Reson. 41 (2011) 1–18.

[49] J.C. Bischof, B. Mahr, J.H. Choi, M. Behling, D. Mewes, Use of X-ray tomographyto map crystalline and amorphous phases in frozen biomaterials, Ann.Biomed. Eng. 35 (2007) 292–304.

[50] K. Morita, Y. Ogawa, C.H.I.N. Thai, F. Tanaka, Soft X-ray image analysis todetect foreign materials in foods, Food Sci. Technol. Res. 9 (2003) 137–141.

[51] M. Shahin, E. Tollner, R. McClendon, H. Arabnia, Apple classification based onsurface bruises using image processing and neural networks, Trans. ASAE 45(2002) 1619–1628.

[52] S. Kim, T. Schatzki, Apple watercore sorting system using X-ray imagery: I.Algorithm development, Trans. ASAE 43 (2000) 1695–1702.

[53] M.A. Shahin, E. Tollner, R. Gitaitis, D. Sumner, B. Maw, Classification of sweetonions based on internal defects using image processing and neural networktechniques, Trans. ASAE 45 (2002) 1613–1618.

[54] R. Mousavi, T. Miri, P.W. Cox, P.J. Fryer, A novel technique for ice crystalvisualization in frozen solids using X-ray micro-computed tomography, J.Food Sci. 70 (2005) e437–e442.

[55] R. Mousavi, T. Miri, P.W. Cox, P.J. Fryer, Imaging food freezing using X-raymicrotomography, Int. J. Food Sci. Technol. 42 (2007) 714–727.

[56] F. Mendoza, P. Verboven, Q.T. Ho, G. Kerckhofs, M. Wevers, B. Nicolaï,Multifractal properties of pore-size distribution in apple tissue using X-rayimaging, J. Food Eng. 99 (2010) 206–215.

[57] H. Kraggerud, J. Wold, M. Hoy, R. Abrahamsen, X-ray images for the control ofeye formation in cheese, Int. J. Dairy Technol. 62 (2009) 147–153.

[58] Y. Tao, Z. Chen, H. Jing, J. Walker, Internal inspection of deboned poultry usingX-ray imaging and adaptive thresholding, Trans. ASAE 44 (2001)1005–1010.

[59] J.S. Kwon, J.M. Lee, W.Y. Kim, Real-time detection of foreign objects using X-ray imaging for dry food manufacturing line, IEEE Int. Symp. Consum.Electron. (2008) 1–4.

[60] D.S. Narvankar, C.B. Singh, D.S. Jayas, N.D.G. White, Assessment of soft X-rayimaging for detection of fungal infection in wheat, Biosyst. Eng. 103 (2009)49–56.

[61] D. Mery, I. Lillo, H. Loebel, V. Riffo, A. Soto, A. Cipriano, J.M. Aguilera,Automated fish bone detection using X-ray imaging, J. Food Eng. 105 (2011)485–492.

[62] B.K. Cho, J.M.K. Irudayaraj, Foreign object and internal disorder detection infood materials using noncontact ultrasound imaging, J. Food Sci. 68 (2003)967–974.

[63] M.J.W. Povey, D.J. McClements, Ultrasonics in food engineering. Part I:introduction and experimental methods, J. Food Eng. 8 (1988) 217–245.

[64] C.H. Frazier, Q. Tian, A. Ozguler, S.A. Morris, W.D. O’Brien Jr, High contrastultrasound images of defects in food package seals, IEEE Trans. Ultrason.Ferroelectr. Freq. Control 47 (2000) 530–539.

[65] N.N. Shah, P.K. Rooney, A. Ozguler, S.A. Morris, W.D. OBrien, A real-timeapproach to detect seal defects in food packages using ultrasonic imaging, J.Food Prot. 64 (2001) 1392–1398.

[66] I. Youssao, V. Verleyen, P. Leroy, Prediction of carcass lean content by real-time ultrasound in Pietrain and negative stress Pietrain, Anim. Sci. 75 (2002)25–32.

[67] W.D. Harron, R. Dony, Predicting quality measures in beef cattle usingultrasound imaging, Comput. Intell. Image Process. (2009) 96–103.

[68] T.H. Gan, D.A. Hutchins, D.R. Billson, Preliminary studies of a novel air-coupled ultrasonic inspection system for food containers, J. Food Eng. 53(2002) 315–323.

[69] T.H. Gan, P. Pallav, D.A. Hutchins, Non-contact ultrasonic qualitymeasurements of food products, J. Food Eng. 77 (2006) 239–247.

[70] E. Hæggström, M. Luukkala, Ultrasound detection and identification offoreign bodies in food products, Food Control 12 (2001) 37–45.

[71] B. Zhao, O.A. Basir, G.S. Mittal, Detection of metal, glass and plastic pieces inbottled beverages using ultrasound, Food Res. Int. 36 (2003) 513–521.

[72] P. Pallav, D.A. Hutchins, T.H. Gan, Air-coupled ultrasonic evaluation of foodmaterials, Ultrasonics 49 (2009) 244–253.

[73] N. Arora, D. Martins, D. Ruggerio, E. Tousimis, A.J. Swistel, M.P. Osborne, R.M.Simmons, Effectiveness of a noninvasive digital infrared thermal imagingsystem in the detection of breast cancer, Am. J. Surg. 196 (2008) 523–526.

[74] A.A. Gowen, B.K. Tiwari, P.J. Cullen, K. McDonnell, C.P. O’Donnell, Applicationsof thermal imaging in food quality and safety assessment, Trends Food Sci.Technol. 21 (2010) 190–200.

[75] J. Rahkonen, H. Jokela, Infrared radiometry for measuring plant leaftemperature during thermal weed control treatment, Biosyst. Eng. 86(2003) 257–266.

[76] D.L. Goedeken, C.H. Tong, R.R. Lentz, Design and calibration of a continuoustemperature measurement system in a microwave cavity by infraredimaging, J. Food Process. Preserv. 15 (1991) 331–337.

[77] J.G. Ibarra, Y. Tao, H. Xin, Combined IR imaging-neural network method forthe estimation of internal temperature in cooked chicken meat, Opt. Eng. 39(2000) 3032–3038.

[78] J. Varith, G. Hyde, A. Baritelle, J. Fellman, T. Sattabongkot, Non-contact bruisedetection in apples by thermal imaging, Innovative Food Sci. Emerg. Technol.4 (2003) 211–218.

[79] R. Vereycken, C. Bravo, H. Ramon, J. De Baerdemaeker, V. Van Linden,Detection technique for tomato bruise damage by thermal imaging, Int. Soc.Horticultural Sci. 599 (2003) 389–394.

[80] M. Fuller, M. Wisniewski, The use of infrared thermal imaging in the study ofice nucleation and freezing of plants1, J. Therm. Biol. 23 (1998) 81–89.

[81] C. Gariepy, J. Amiot, S. Nadai, Ante-mortem detection of PSE and DFD byinfrared thermography of pigs before stunning, Meat Sci. 25 (1989) 37–41.

[82] L. Nanni Costa, C. Stelletta, C. Cannizzo, M. Gianesella, D. Lo Fiego, M.Morgante, The use of thermography on the slaughter-line for the assessmentof pork and raw ham quality, Ital. J. Anim. Sci. 6 (2010) 704–706.

274 Q. Chen et al. / Trends in Analytical Chemistry 52 (2013) 261–274

[83] S. Wang, J. Tang, T. Sun, E. Mitcham, T. Koral, S. Birla, Considerations in designof commercial radio frequency treatments for postharvest pest control in in-shell walnuts, J. Food Eng. 77 (2006) 304–312.

[84] P. Fito, M. Ortolá, R. De los Reyes, P. Fito, E. De los Reyes, Control of citrussurface drying by image analysis of infrared thermography, J. Food Eng. 61(2004) 287–290.

[85] A. Manickavasagan, D. Jayas, N. White, F. Jian, Thermal imaging of a storedgrain silo to detect a hot spot, Appl. Eng. Agric. 22 (2006) 891.

[86] A. Manickavasagan, D. Jayas, N. White, Thermal imaging to detect infestationby Cryptolestes ferrugineus inside wheat kernels, J. Stored Prod. Res. 44 (2008)186–192.

[87] P. Meinlschmidt, V. Maergner, Detection of foreign substances in food usingthermography, Int. Soc. Opt. Photonics 4710 (2002) 565–571.

[88] G. Ginesu, D.D. Giusto, V. Margner, P. Meinlschmidt, Detection of foreignbodies in food by thermal image processing, IEEE Trans. Ind. Electron. 51(2004) 480–490.

[89] W. Jun, M.S. Kim, K. Lee, P. Millner, K. Chao, Assessment of bacterial biofilmon stainless steel by hyperspectral fluorescence imaging, Sens. Instrum. FoodQual. Saf. 3 (2009) 41–48.

[90] M. Kokawa, K. Fujita, J. Sugiyama, M. Tsuta, M. Shibata, T. Araki, H. Nabetani,Visualization of gluten and starch distributions in dough by fluorescencefingerprint imaging, Biosci. Biotechnol. Biochem. 75 (2011) 2112–2118.

[91] M. Kokawa, K. Fujita, J. Sugiyama, M. Tsuta, M. Shibata, T. Araki, H. Nabetani,Quantification of the distributions of gluten, starch and air bubbles in doughat different mixing stages by fluorescence fingerprint imaging, J. Cereal Sci. 55(2012) 15–21.

[92] A.A. Adedeji, L. Liu, M.O. Ngadi, Microstructural evaluation of deep-fat friedchicken nugget batter coating using confocal laser scanning microscopy, J.Food Eng. 102 (2011) 49–57.

[93] C. Byoung-Kwan, M.S. Kim, K. Chao, A.M. Lefcourt, K. Lawrence, B. Park,Detection of fecal residue on poultry carcasses by laser-induced fluorescenceimaging, J. Food Sci. 74 (2009) E154–E159.

[94] C.-C. Yang, M.S. Kim, S. Kang, B.-K. Cho, K. Chao, A.M. Lefcourt, D.E. Chan, Redto far-red multispectral fluorescence image fusion for detection of fecalcontamination on apples, J. Food Eng. 108 (2012) 312–319.

[95] W. Jun, M.S. Kim, B.-K. Cho, P.D. Millner, K. Chao, D.E. Chan, Microbial biofilmdetection on food contact surfaces by macro-scale fluorescence imaging, J.Food Eng. 99 (2010) 314–322.

[96] J. Qin, K. Chao, M.S. Kim, S. Kang, B.K. Cho, W. Jun, Detection of organicresidues on poultry processing equipment surfaces by LED-inducedfluorescence imaging, Appl. Eng. Agric. 27 (2011) 153–161.

[97] N.A. Rakow, K.S. Suslick, A colorimetric sensor array for odour visualization,Nature 406 (2000) 710–713.

[98] X. Huang, J. Xin, J. Zhao, A novel technique for rapid evaluation of fishfreshness using colorimetric sensor array, J. Food Eng. 105 (2011) 632–637.

[99] X. Luo, P. Liu, C. Hou, D. Huo, J. Dong, H. Fa, M. Yang, A novel chemicaldetector using colorimetric sensor array and pattern recognition methods forthe concentration analysis of NH3, Rev. Sci. Instrum. 81 (2010). 105113–1051136-6.

[100] C.J. Hou, C. Zhang, K. Suslick, Analysis and evaluation of teas by colorimetricsensor array in both liquid and gas phases, Abstr. Papers Am. Chem. Soc. 230(2005) 261–262.

[101] C. Zhang, K.S. Suslick, Colorimetric sensor array for soft drink analysis, J.Agric. Food Chem. 55 (2006) 237–242.

[102] S.A. Chan, J.S. Lee, Y.T. Chang, Colorimetric sensor array for qualitative wateranalysis, Aust. J. Chem. 62 (2009) 1040–1046.

[103] B.A. Suslick, L. Feng, K.S. Suslick, Discrimination of complex mixtures by acolorimetric sensor array: coffee aromas, Anal. Chem. 82 (2010) 2067–2073.

[104] C. Zhang, D.P. Bailey, K.S. Suslick, Colorimetric sensor arrays for the analysisof beers: a feasibility study, J. Agric. Food Chem. 54 (2006) 4925–4931.

[105] Z. Ya, K. He, Z. Lu, B. Yi, C. Hou, S. Shan, D. Huo, X. Luo, Colorimetric artificialnose for baijiu identification, Flavour Fragance J. 27 (2012) 165–170.

[106] H. Qin, D. Huo, L. Zhang, L. Yang, S. Zhang, M. Yang, C. Shen, C. Hou,Colorimetric artificial nose for identification of Chinese liquor with differentgeographic origins, Food Res. Int. 45 (2012) 45–51.

[107] D.Q. Huo, M.M. Yin, C.J. Hou, Identification of different aromatic Chineseliquors by colorimetric array sensor technology, Chin. J. Anal. Chem. 39(2011) 516–520.

[108] Q. Ouyang, J. Zhao, Q. Chen, H. Lin, Classification of rice wine according todifferent marked ages using a novel artificial olfactory technique based oncolorimetric sensor array, Food Chem. 138 (2013) 1320–1324.

[109] Y. Salinas, J.V. Ros-Lis, J.L. Vivancos, R. Martinez-Manez, M.D. Marcos, S.Aucejo, N. Herranz, I. Lorente, Monitoring of chicken meat freshness bymeans of a colorimetric sensor array, Analyst 137 (2012) 3635–3643.

[110] S.H. Lim, C.J. Musto, E. Park, W. Zhong, K.S. Suslick, A colorimetric sensorarray for detection and identification of sugars, Org. Lett. 10 (2008) 4405–4408.

[111] Q. Chen, A. Liu, J. Zhao, Q. Ouyang, Z. Sun, L. Huang, Monitoring vinegar aceticfermentation using a colorimetric sensor array, Sens. Actuators, B 183 (2013)608–616.

[112] H.-M. Lai, S.-C. Hwang, Water status of cooked white salted noodlesevaluated by MRI, Food Res. Int. 37 (2004) 957–966.

[113] Q. Chen, A. Liu, J. Zhao, Q. Ouyang, Classification of tea category using aportable electronic nose based on an odour imaging sensor array, J. Pharm.Biomed. Anal. 84 (2013) 77–83.