Adam was showing us yesterday at the MIT-Knight Civic Media conference that you can use GeniePro to 'train' a computer to identify different types of land cover. In this picture, he trained it to separate 7 zones, including roof, grass, sand, and pavement. I believe GeniePro was originally written by Los Alamos National Labs, and should be public domain code -- the version we're using is a commercial variant by Observera.
Adam sez he'll be posting a video soon (please!) but this has a lot of promise for, say, classifying wetlands areas, identifying species, etc. and it's never really been done on imagery of our resolution -- the above is a 6cm/px image Adam took a couple weeks ago at the University of South Carolina Beaufort.
Ned Horning points out that Fiji has a similar open source plugin: http://fiji.sc/Trainable_Weka_Segmentation
Reply to this comment...
Log in to comment
Login to comment.