THEMIS Visual Images and Slope Classifications

1
THEMIS Visual Images and Slope Classifications

• Katie Bryson and Kate Coleman
• Remote Sensing Project
• Spring 2007

2
Mars Odysseys THEMIS

• Thermal Emission Imaging System
• Spectral resolution
• 5 visual bands
• 0.425, 0.540, 0.654, 0.749, 0.860 µm
• 10 infrared bands
• 6.78 (used twice), 7.93, 8.56, 9.35, 10.21,
  11.04, 11.79, 12.57, 14.88 µm
• Spatial resolution
• Visual - 18 m
• IR - 100 m

3
THEMIS

• Detectors
• Visual - 1,024 x 1,024 element silicon array
• IR - 320 x 240 element uncooled microbolometer
  array
• Design
• Visual - strip filter frame-scan
• IR - strip filter pushbroom

Representation of the Visible Detector - A single
CCD array is used. The five color filters are
bonded to the detector in strips as shown here.
Simulation of the motion of the Visible Detector
THEMIS IR focal-plane stripe-filter layout
4
Conceptual Model
5
Image Acquisition
6
Initial Image

• File Formats
• IMG
• not Imagine IMG
• Cube
• raw and calibrated data
• TIFF
• Chosen because of ease of use
• Visual band 3 (0.654 µm)
• Shows surface albedo with high contrast and
  signal/noise ratio
• Distinct Slopes

7
Unsupervised Classification

• Initial - 20 classes, max 20 iterations
• Too many classes to distinguish slopes
• Final - 10 classes, max 20 iterations

8
Unsupervised Classification 20 and 15 classes
20
15
9
Unsupervised Classification 10 classes
10
Supervised Classification

• 9 classes total

11
Applying Classifications

• Dissimilar to original image
• Distinctly different slopes
• Large regions of mostly flat

12
Unsupervised Classification 10 classes
Applying the unsupervised classes as a supervised
classification
Unsupervised classification using previous
unsupervised signature means
13
Supervised Classification
14
Accuracy Assessment

• Lack ground reference data
• Expected map accuracy 85, acceptable error 10
• Sample size 51
• Error likely greater than 10
• No higher spatial resolution image (MRO)
• No easy access to MOLA data

15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
Mars Orbital Camera Image
Flat Base High Slope in Shadow High Bright
Slope Low Bright Slope High Slope in Darkest
Shadow
Low Slope Mid Slope Bright Mid Slope Shadow Light
Side Slope Base
25
Conclusions

• NASA has many of its own file formats
• Supervised classification worked best for
  different images
• Need for other data sets to determine accuracy
• Ground reference data is not available for most
  of the planet
• Ability to apply the classification to other
  similar data sets
• It is possible to classify different slopes