multispectral camera

Multispectral Camera

Simon Belkin, Audrey Finken, Grant George, & Matthew Walczak

Final Project Review

Team Parente

Problem & Solution• Problem - Scientists desire the

ability to capture measurements based on reflected energy (radiance)

• Common camera technology only captures light in three primary wavelengths: Red, Green, and Blue (RGB) in the visible spectrum

Solution – Multispectral camera captures wavelengths throughout the entire spectrum (including near-infrared/infrared)

System Design

Final Design Review Deliverables

Filter wheel turning and stopping at each filter for period of time, capturing image, and rotating to the next filter

Aberration mitigation for all filters Transmitting captured image data from camera to

Raspberry Pi Identifying geological markers present in rocks and

soil

Filter Wheel Mechanics

Filter Wheel Mechanics FDR Solution

Program the filter wheel location in Python Replace Hall Effect Sensor and add to Python program to

identify start location Accomplished

A program turns the filter wheel. Hall Effect sensor used to recognizes start point location. With the start point location the program can turn to

specific filter and run through the filters pausing for outside response.

Filter SelectionSelected• 436nm• 670nm• 750nm• 860nm• 990nmOther• 425nm• 450nm• 510nm

Optical System

Optical System Out of focus condition due to the addition of filter wheel

required compensation. 3.75” (95.3mm) of extension tubing. A +33mm convex relay lens. Thin lens equation, 1/di = 1/f – 1/do, determines the

distance to the object, do, and to the image, di.

Required Focal Length Adjustment Cameras are calibrated to focus on the visible

portion of the light spectrum. As you travel further away from the visible spectrum and into the infrared spectrum, a slight shift in focus is necessary for camera to obtain sharpness of image.

Filter Position at Light PathCurrent Arrangement of Filter Wavelength

(nm)Dimension Needed for Focus Correction

(mm)1 Blank 9.02 425 8.53 436 8.54 450 NA5 510 9.66 will be 670nm NA7 750 10.98 860 11.09 990 11.5

10 Blank 9.0

Adjusted Lens Position (between item #2 & #3)

Pi to Camera Interface

Pi to Camera InterfaceProblem – Driver Installation Drivers provided, not officially “supported” “Fedora 5” build conversion to Raspbian “wheezy” Attempt to Cross-compile module Module compiled, but unrecognizable by OS.

Cannot start via terminal command Pi integration determined to be low priority, efforts

focused on ensuring quality image capture

Image Processing

Solution – Image Registration

Problem - aberration An optical aberration is a deviation from the optical

predictionsSolution- Image registration Image Registration is the process of estimating an optimal

transformation between two images for all filters Use of checkerboard test in Matlab to calibrate camera

Images Focused Through all Filters

Filter 2 (425nm – Visible Spectrum) Filter 3 (436nm – Visible Spectrum)

Filter 5 (510nm – Visible Spectrum) Filter 7 (750nm – Infrared Spectrum)

Filter 8 (860nm – Infrared Spectrum) Filter 9 (990nm – Infrared Spectrum)

Sift matching points on images Points from image matching points from two

images

Images after image registration

Questions