6. image classification objectives of...
TRANSCRIPT
1
6. Image Classification6. Image Classification6.1 Concept of Classification6.1 Concept of Classification
Objectives of ClassificationAdvantages of Multi-Spectral data for ClassificationVariation of Multi-Spectra DataSegmentation in Feature DomainSupervised and Un-Supervised CalssificationLand Cover and Land UseExisting Land Cover Class
Objectives of ClassificationObjectives of ClassificationTo create Maps such as Landuse Map, Forest Map, Crop Map, Shrimp pond Map, Mangrove Map, etc.Carry out quantitative interpretation using mathematical / statistical modeling.To assign corresponding class to groups with homogeneous characteristics, with the aim of discriminating multiple objects from each other within the image. The level is called class. Classification will be executed on the base of spectrally defined features, such as density, (texture etc. ) in the feature space. It can be said that classification divides the feature space into several classes based on a decision rule. Classes are for such as Land use, Land Cover, Crop Type, Forest Types, and etc.
RS Image ClassificationRS Image ClassificationMulti-Spectral Data Classification
Assumption - Different surface materials have defferent sepectral reflectanceK-dimensional vector ( K:number of band )divide K-dimensional feature space into few regions ( classes )
Concept of Classification of Remote SensingConcept of Classification of Remote Sensing
Segmentation in Feature DomainSegmentation in Feature Domain
In general, the separation of all classes requires more than twospectral bands. Because the clusters occur in K-dimensions.
0.4 0.6 0.8 1.21.0 1.4 1.6 1.8 2.0 2.2 2.4 2.6
20
40
50
60
70
80
10
0
Vegetation
Soil
Clear River Water Turbid River Water
Wavelength (µm)
Perc
ent R
efle
ctan
ce
30
Spectral ReflectanceSpectral Reflectance
2
MultiMulti--spectral Classificationspectral ClassificationThe spectral signature is a K-d imensional vector whose coordinates are the measured radiance in each spectral band. If every pixel from each land cover has same rad iance with in the class, only 1 band (IR) would be enough for classification for the case of water, soil and vegetation below.
NIR
VR
Variation of Variation of MultispectralMultispectral datadataIn reality, the spectral radiance of a given surface material is not characterized by a single, deterministic curve, but by a family of curves with a range of variability.
NIR
NIR
Segmentation in MultiSegmentation in Multi--dimensional feature spacedimensional feature spaceThus, it is very common to find big overlaps among distributions in one band informat ion.By combin ing other bands, we can improve the accuracy of classification, which is a segmentation in a mult i-dimensional feature space.
Supervised and UnSupervised and Un--Supervised ClassificationSupervised ClassificationSupervised Classification
Classify each pixel into a pre-established class.Population statistics of each class is to be identified by training areas.Each pixel will be classified into a class which has similar ( nearest ) property with the pixel.
Un-supervised ClassificationAnalyze inherent structure of the dataUnconstrained by external knowledge about areaWhen knowledge about the area is not enough
CombinationUn-Supervised Classification -> Ground Truth -> Supervised Classification
Land Cover and Land UseLand Cover and Land UseLand Cover means materials which covers the ground. ( Soil )Land use is how land is being used. ( Play ground, Harvested Paddy )A land use class consists of some land cover classes.
Urban Area ( land use ) consists of concrete, vegetation, bare land, etc..
Generally, image classification gives land cover, but not land use, because what RS is observing is spectral signature of object ( material ).
Existing Land Cover Existing Land Cover Classification SystemClassification System
There are several organizations/groups such as IGBP, UNEP/FAO, USGS, which are trying to develop land cover data set of global or continental area. The Land Cover Working Group(LCWG) of the Asian Association on Remote Sensing(AARS) also aims to develop global or continental land cover data set. It is a global land cover data set but focuses on Asian and Oceania regions in terms of ground truth.Appropriate Land classification system ( class ) are developed according to the project’s objectives
3
Global Ecosystems Legend Value DescriptionGlobal Ecosystems Legend Value DescriptionGlobal Ecosystems (Olson, 1994a, 1994b) Global Ecosystems (Olson, 1994a, 1994b)
Global Ecosystems Legend Value Description 1 Urban 2 Low Sparse Grassland 3 Coniferous Forest 4 Deciduous Conifer Forest 5 Deciduous Broadleaf Forest 6 Evergreen Broadleaf Forests 7 Tall Grasses and Shrubs 8 Bare Desert 9 Upland Tundra 10 Irrigated Grassland 11 Semi Desert 12 Glacier Ice 13 Wooded Wet Swamp 14 Inland Water 15 Sea Water 16 Shrub Evergreen…..
USGSUSGSUSGS Land Use/Land Cover System Legend (Modified Level 2) Value
Code Description (http://landcover.usgs.gov/, http://edcdaac.usgs.gov/glcc/globdoc2_0.html#app1)
Level 1 Class Value Code Description1 Urban or Buil-up Land
1 100 Urban and Built-Up Land
2 Agricultural Land2 211 Dryland Cropland and Pasture 3 212 Irrigated Cropland and Pasture 4 213 Mixed Dryland/Irrigated Cropland and Pasture 5 280 Cropland/Grassland Mosaic 6 290 Cropland/Woodland Mosaic
3 Rangeland 7 311 Grassland 8 321 Shrubland9 330 Mixed Shrubland/Grassland 10 332 Savanna
LCWG, AARSLCWG, AARSThe Land Cover Working Group(LCW G) of the Asian Association on Remote Sensing(AARS) also aims to develop global or continental land cover data set. It is a global land cover data set but focuses on Asian and Oceania regions in terms of ground truth.
Development of Land Development of Land Classification SystemClassification System
Considering objectives of the classification, available data, cost, technical feasibility, we have to develop a classification system.– Agriculture monitoring ?– Urban environment ?– Flood modeling ?– Flood damage assessment ?– Forestry ?
6.2 Ground Truth6.2 Ground TruthGround truth is simply observations or measurements made at or near the surface of the earth in support of an air or space-based remote sensing survey. The location will be acquired by GPS to identify the location onRS imageTo understand the real situation and phenomena on the groundTo provide reference data for– Classification
Development of Classification ProcedureEstimation of statistical property of the class
– VerificationAccuracy Assessment
Ground truth may consist of several types of data acquired before, during, and after an image acquisition. Such measurements and observations may include, but are not limited to:
4
Measurement and observationMeasurement and observation
GPS, algae sampling, and quadratmapping in Yaquina Bay, Oregon
spectral measurements of grasses being made with two separate spectrometers
Remote Sensing for Rice Yield Estimation Remote Sensing for Rice Yield Estimation in Indonesiain Indonesia
Rice main growth stages
Heading
Ripening Harvest
Planting
OPTICAL SATELLITE DATA
BlanakanPamanukan
Ciasem
Patokbeusi
Pusakanagara
Binong
Blanakan
Pamanukan
Ciasem
Patokbeusi
Pusakanagara
Binong
SPOT DATA1999/08/05
LANDSAT-TM DATA1997/07/28
Heading
Harvested
Planting
Taking Spectrophotometer readings
ResultsModel Developments
•Develop a model to find the correlation between LAI and NDVI derived from spectrophotometer readings
LAI Vs. NDVI (Spectophotometer readings)y = 0.233Ln(x) + 0.1829
R2 = 0.7388
0.0
0.1
0.2
0.3
0.40.5
0.6
0.7
0.8
0.9
1.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Leaf Area Index
ND
VI (
(NIR
-IR)/(
NIR
+IR
))
Rice Growth Monitoring UsingRice Growth Monitoring UsingNear Real Time RADARSAT Fine Beam SAR Near Real Time RADARSAT Fine Beam SAR
DataData in in PathumthaniPathumthani
Deployment sites ofcorner reflectors, plottedOn ADEOS AVNIR image
Canada – Japan – ThailandWithin 8 hrs after reception
5
Field locations of reflectors andField locations of reflectors andcorresponding views in the corresponding views in the
imageimage(11a) St ation No. 1
Reflector: 8.53 dBBackground: -21.22 dB
(11b) Station No. 2
Reflector: 9.89 dBBackground: -2.19 dB
(11c) St ation No. 3
Reflector: 9.48 dBBackground: -3.14 dB
(11d) Station No. 4
Reflector: 9.21 dBBackground: -5.08 dB
for geometric correction of RADARSAT
image
For Better Overlay of
Radar Image and Field
Survey Result
Early Vegetative GrowthEarly Vegetative Growth
Longitude1562832 N
GPS:Lat itude678889 E
Location: Klong 3Land Cover: Early vegetative Growth
Time: 15:45Date: 9/03/2003
Early Reproductive GrowthEarly Reproductive Growth
Longitude1557329 N
GPS:Lat itude683070 E
Location: Klong 3Land Cover: early reproductive Growth
Time: 15:55Date: 9/03/2003Steps for Supervised ClassificationSteps for Supervised Classification
In order to determine a decision rule for classification, it is necessary to know the spectral characteristics or features with respect to the population of each class. Due to atmospheric effects, direct use of spectral features measured on the ground are not always available. Sampling of training data from clearly identified training areas, corresponding to defined classes is usually made for estimating the population statistics. Statistically unbiased sampling of training data should be made in order to represent the population correctly.
6.3. Supervised Classification6.3. Supervised Classification
Sampling of Training DataSampling of Training DataGet statistical characteristics for each class
A representative area for each desired class must be located by an analyst.
analyst knowledge, field survey, aerial photographs, existing mapsSelect homogeneous area
Div ide the class into several homogeneous sub-class
Select sufficient number of pixels to estimate class statisticsUnbiased selection of area
Estimation of Population StatisticEstimation of Population Statistic
6
ClassifierClassifierThere are a LOT of classifier algorithms.Such as– Parallelepiped Classifier– Decision Tree Classifier– Minimum distance Classifier– Maximum likelihood Classifier
ParallelpipedParallelpiped ClassifierClassifier
The minimum and maximum DNs for each class are determined and are used as thresholds for classifying the image.Benefits: simple to train and use, computationally fastDrawbacks: pixels in the gaps between the parallelepipes can not be classified; pixels in the region of overlapping parallelepipes can not be classified.
ParallepipedParallepiped ClassifierClassifier Decision Tree ClassifierDecision Tree ClassifierThe decision tree classifier is an hierarchically based classifier which compares the data with a range of properly selected features. The selection of features is determined from an assessment of the spectral distributions or separability of the classes. There is no generally established procedure. Therefore each decision tree or set of rules should be designed by an expert. When a decision tree provides only two outcomes at each stage, the classifier is called a binary decision tree classifier (BDT).
Decision Tree ClassifierDecision Tree ClassifierMinimum Distance ClassifierMinimum Distance Classifier
A “centroid” for each class is determined from the data by calculating the mean value by band for each class. For each image pixel, the distance in n-dimensional distance to each of these centroids is calculated, and the closest centroid determines the class.Benefits: mathematically simple and computationally efficient Drawback: insensitive to different degrees of variance in spectral response data.
7
Minimum Distance ClassifierMinimum Distance Classifier Maximum Likelihood ClassifierMaximum Likelihood ClassifierMost Popular methodsMaximum likelihood classification uses mean and variance-covariance in class spectra to determine classification scheme. It assumes that the spectral responses for a given class has normal distribution.A pixel with the maximum likelihood is classified into the corresponding class
BayesBayes TheoryTheoryfeature x --- for example, the gray level of each pixel
p( x | i ) : probability density function in class ip( i ) : a priori probabilitiesp( i | x ) : a posteriori probabilities
Bayes Rulep( i | x ) : p( x | i ) p( i ) / p( x )
If we observed feature x, what is the probability to be class i ?p(x ) = p ( x | i ) p( i )
Bayes Dicision Ruleone dimensional, two-class classification problema pixel belongs to class 1 if p(x|1)p(1) > p(x|2)p(2)a pixel belongs to class 2 if p(x|2)p(2) > p(x|1)p(1)
BayesBayes Decision RuleDecision Rulep(forest) = 0.6p(Agr) = 0.4
P(f1|Forest)=0.3P(f2|Forest)=0.7P(f1|Agr)=0.9P(f2|Agr)=0.1
p(f1|Forest) *p(Forest)= 0.3 x 0.6 = 0.18
p(f2|Forest) *p(Forest)= 0.7 x 0.6 = 0.42
p(f1|Agr) *p(Agr)= 0.9 x 0.4 = 0.36
p(f2|Agr) *p(Agr)= 0.1 x 0.4 = 0.04
p(Forest|f1)=p(f1|Forest)*p(Forest) / p(f1) = 0.18 / 0.54 = 0.33p(Agr|f1)=p(f1|Agr)*p(Agr) / p(f1) = 0.36 / 0.54 = 0.67
p(Forest|f2)=p(f2|Forest)*p(Forest) / p(f2) = 0.42 / 0.46 = 0.91
p(Agr|f2)=p(f2|Agr)*p(Agr) / p(f2) = 0.04 / 0.46 = 0.09
f2f1 f2f10.10.90.70.3
Forest
0.6
Agriculture
0.4
AgricultureForest
.040.360.420.18
AgForestAgriForest
f2f10.54 0.46
0.18 0.36 0.42 0.04
0.33 0.67 0.91 0.09
DiscriminantDiscriminant FunctionFunctionThe Bayes Dicision Rule is restated as
a pixel belongs to class 1 if D1(x) > D2(x)a pixel belongs to class 2 if D2(x) > D1(x)
where Di is called discriminant function and is given byDi(x) = p( x | i ) p( i )
However P(i) is unknown, we assume p(i)=p(j)
Assumption of Normal Assumption of Normal DistributionDistribution
p( x | i ) = 1
2 2 exp ( –( x – i )
2
2 i2 )
i = mean of x for classii2 = varianceof x for classi
If the class probability distributions are normal
Bayes optimal discriminant function for class i is then Di(x) = ln [ p( x | i ) p( i ) ]
= ln [ p( i ) ] – 12 ln [ 2 ] – 1
2 ln [ i2 ] –
( x – i )2
2 i2
Di(x) = – 1
2 ln [ i2 ] –
( x – i ) 2
2 i2
p(i) is unknown. Assumption of p(i) = p(j),
8
Extension to K DimensionExtension to K Dimension
p( x | i ) = 1
( 2 )K / 2i
1/2exp[ – 12 ( X– i )t i
–1 ( X – i ) ]
Di(x) = ln [ p( i ) ] – K2 ln [ 2 ] – 1
2 ln [ i ] – 12 ( X– i )t
i–1 ( X – i )
Di(x) = – 12 ln [ i ] – 1
2 ( X – i )t
i–1 ( X– i )
X : vectorof imagedata( K dimension) X= [ x1 , x2 , ...,xk ]i: meanvectorfor classi i = [ m1 , m2 , ...,mk ]
i: variance–covariancematrixfor class i
i =
11 11 1k
21 22 2k
k1 k2 kk
i : determinantof i
i–1: inversematrixof i
ThresholdingThresholdingEliminate pixels which have low posteriori probability
Actual DistributionActual Distribution 6.4 Unsupervised Classification6.4 Unsupervised ClassificationTo determine the inherent structure of the data, unconstrained by external knowledge about the area.To produce clusters automatically, which consists of pixels withsimilar spectral signature
Hierarchical ClusteringEvaluate distance between clustersMerge a pair of clusters which have the minimum distance. Members are not reallocated to different clusters
Non-Hierarchical ClusteringK-mean, ISODATA methodReallocation of membersMerge and Division of clusters
Hierarchical clusteringHierarchical clustering ISODATA methodISODATA method
9
ISODATA Unsupervised Classification ISODATA Unsupervised Classification exampleexample
Allocation of Land Cover/Use to ClustersAllocation of Land Cover/Use to ClustersUn-supervised classification gives only clusters, without any interpretation; land cove/useOperator must give class name to each clusters.
Usual ProcessUn-Supervised Classification before field survey.One land cover/use might have been divided into several clusters.More than two land cover/use classes might have been merged in one cluster.Visit area where operator cannot identify land use/cover.Based on above, carry out supervised classification
6.5 Accuracy Assessment6.5 Accuracy AssessmentAccuracy assessments determine the quality of the information derived from remotely sensed data (Congaltonand Green, 1999).Accuracy assessment is important to produce reliable maps.Assessments can be either qualitative or quantitative. In qualitative assessments, we determine if a map "looks right" by comparing what we see in the imagery with what we see on the ground. However quantitative assessments attempt to identify and measure remote sensing-based map error. In such assessments, we compare map data with reference or ground truth data.
Reference/Ground truth Reference/Ground truth data collectiondata collection
Usually we divide ground truth into two.– 50% is used for supervised classification training– 50% is used for accuracy assessment
Aerial photographsOther MapsGround based data is assumed to be 100% correct in accuracy assessments, hence it's very important that the data is collected carefully. It should be collected consistently with vigilant quality control.
Common quantitative error Common quantitative error assessmentsassessments
Error Matrix or Confusion Matrix - assesses accuracy for each class as well as for the whole image; this includes errors of inclusion and errors of exclusion
We must accept some level of error as a trade off for the cost savings of remotely sensed data (Congalton and Kass, 1999)
Confusion MatrixConfusion MatrixUrban Crop Range Water Forest Barren Total PA EO
Urban 150 21 9 7 17 30 234 64.1% 35.9%Crop 0 730 93 14 115 21 973 75.0% 25.0%
Range 33 121 320 23 54 43 594 53.9% 46.1%Water 3 18 11 83 8 3 126 65.9% 34.1%Forest 23 81 12 4 350 13 483 72.5% 27.5%
Barren 39 8 15 3 11 115 191 60.2% 39.8%Total 248 979 460 134 555 225 1748
CA 60.5% 74.6% 69.6% 61.9% 63.1% 51.1%EC 39.5% 25.4% 30.4% 38.1% 36.9% 48.9%
Total Pixel 2601Correct Pixel 150+730+320+83+350+115 = 1748Overall Accuracy =1748/2601 67.2%
PA Producers AccuracyCA(UA) Consumer's (User's) Accuracy
EO Error of Omission = 100%-PAEC Error of Commission = 100%-CA
Classified
Refe
renc
e (
Gro
und
Tru
th )
10
ProducerProducer’’s accuracys accuracy
Probability of a reference pixel being correctly classified.How well has a certain type been classified?
Error of omission– error of omission (%) = 100 % - Prod’s acc. (%)– Proportion of observed features on ground that are not
classified.
UserUser’’s accuracys accuracyoror
ConsumerConsumer’’s Accuracys AccuracyProbability that a pixel classified on the image actually represent that category on the ground.Reliability of map from user’s view
Error of commission– error of commission (%) = 100 % - User’s acc. (%)
6.6 Post 6.6 Post -- ClassificationClassification– Smoothing for Classified Image
Noisy appearance - isolated pixels, small groups of pixelsClass-homogeneous regions is importantSmoothing algorithm– minimum area constraint– majority filter
– Import to GIS and Publish as a map
Smoothing with minimum area constraintSmoothing with minimum area constraint– Alters those class-homogeneous regions that have less
than specified minimum area to the class that is the majority along the original boundary.
The yellow area is only 4 pixelsBoundary with A:6, B:5, D:3 -> Replace with A
A A A A B
A C C C B
A C B B B
D D D D D
A A A A B
A A A A B
A A B B B
D D D D D
Smoothing with majority filterSmoothing with majority filter– If there is a majority class in 3x3 window, replace
central pixel with the majority class
A A A
A C B
C A A
Class # Pixels
A 6
B 1
C 2
A A A
A A B
C A A
A A B
C C B
C B A
Class # Pixels
A 3
B 3
C 3
A A B
C C B
C B A
No majority !No Change
Class A is themajority !Replace Central Pixel to A
Import Classification to GISImport Classification to GISThe result of image classification is in raster format that will be transformed to vector shape file format for data input of a GIS.
11
Import Classification to GISImport Classification to GIS Map Development ExampleMap Development Example
Phi Ph i Coral Reef mapLive Cora l on Substratum zone
Dead Reef zone
Reef Dense zone
Dead Reef flat zone
Reef on Substratum zone
Coral Rock zone
Substra ta S andy zone
Substra ta Rubble zone
Substra ta Rock zone
Substra ta S ilt zone
L and and deep wate r area
Habitat type
Habitat type map of Phi Phi Don Island.
Map Development ExampleMap Development Example Map Development ExampleMap Development Example
End of Lecture !End of Lecture !