Madagascar2011 - 09 OTB Change detection framework
0 likes1,264 views
This document discusses change detection techniques using the Orfeo Toolbox. It presents several use cases for identifying changes between images using modules for change detection, SVM classification, and band math. The main challenges are distinguishing exceptional changes from natural changes over time in vegetation. Menus and steps are provided to open and preprocess images, classify changes versus unchanged areas, learn from sample inputs, and refine results.
Madagascar2011 - 09 OTB Change detection framework
- 1. Orfeo Toolbox Change detection Stéphane MAY stephane.may@cnes.fr orfeo-toolbox.org 1
- 2. Change detection Goal Identify changes between 2 images Main problems Exceptional change (crisis) mixed with Natural changes (vegetation evolution) orfeo-toolbox.org 2
- 3. Monteverdi Filtering Change Detection orfeo-toolbox.org 3
- 4. Use case 1 : change detection module Menu File > Open ./img_gloucester_spot_av_ap/Spot5-Gloucester-before.tif ./img_gloucester_spot_av_ap/Spot5-Gloucester-after.tif Menu > Filtering > Change detection Click on changed class ➢ Select changed areas Click on unchanged class ➢ Select unchanged areas Learn Display results orfeo-toolbox.org 4
- 5. Use case 2 : change detection with SVM classif. module Menu File > Open ./img_gloucester_spot_av_ap/Spot5-Gloucester-before.tif ./img_gloucester_spot_av_ap/Spot5-Gloucester-after.tif Menu File > Concatenate image Insert “after channels” first Menu Learning > SVM classification Create changed class Create unchanged class Learn Display Add new areas to refine your results Option : add features to enhance results (Feature Extraction Module) orfeo-toolbox.org 5
- 6. Use case 3 : RADAR change detect. with SVM classif. Menu File > Open ./img_gloucester_av_ap/ERSAvantSmall.png ./img_gloucester_av_ap/ERSApresSmall.png Menu SAR >Despeckle Lee filter ➢ Radius 5 Menu File > Concatenate image Insert crisis image first. Insert before image twice for visualisation. Menu Learning > SVM classification Create changed class Create unchanged class Learn, Display Add new areas to refine your results orfeo-toolbox.org 6
- 7. Use case 4 : RADAR change detection with band math Menu File > Open ./img_gloucester_spot_av_ap/Spot5-Gloucester-before.tif ./img_gloucester_spot_av_ap/Spot5-Gloucester-after.tif Menu SAR >Despeckle Lee filter ➢ Radius 5 Menu Filtering > Band math abs(im1b1-im2b1) min(im1b1/im2b1, im2b1/im1b1) Invent your formula ! orfeo-toolbox.org 7
- 8. Use case 5 : classif. based change detection Menu File > Open ./output_fus1.tif ./output_fus2b.tif Menu Learning > SVM classification (file1) Create land cover classes (water, vegetation, soil, roads, houses) Menu Learning > SVM classification (file2) Create land cover classes (water, vegetation, soil, roads, houses) Menu File > Concatenate images (land cover files) Menu Learning > SVM classification Create changed class / unchanged class Learn, Display Menu filtering > Band math orfeo-toolbox.org Select interesting couples of values from land cover files 8
- 9. Monteverdi Thank you for your attention ! orfeo-toolbox.org 9