rahdhitya Posted February 6, 2013 Report Share Posted February 6, 2013 Hello Friends, Currently, i have some researches about object classification. I am using ecogniton before, but i am hardly knowing how to use DEM as a thematic layer on that. How to add altitude information from DEM into our ruleset? Is there anything feature represent altiltude (DEM) ? I also want to use CART analysis as decision tree for choosing some features to split all my class. But, i dont know how to use it. I ask the author from paper i read, he told in ecognition 8, CART already built in so i dont need to use another software. Anyone can help me how to use that? Thanks for any help friend. Quote Link to comment Share on other sites More sharing options...
3dbu Posted February 6, 2013 Report Share Posted February 6, 2013 is nice to use the ecognition but i usually open it and play a little bit ....did you search in other place about this tipic, regards Quote Link to comment Share on other sites More sharing options...
rahdhitya Posted February 7, 2013 Author Report Share Posted February 7, 2013 Yes, i already used eCognition more than a year, but i didn't know that is CART algorithm built-in inside. I also had search a lot papers and tutorials about those but i didn't help much. Any solutions friends? Quote Link to comment Share on other sites More sharing options...
rahmansunbeam Posted February 7, 2013 Report Share Posted February 7, 2013 umm, never heard about CART analysis, but did some googling and found that there is a standalone software for that. Try this paper also. BTW, I did find few articles you can use. https://docs.google.com/viewer?a=v&q=cache:E2oQ16KX5IgJ:remotesensing.montana.edu/pdfs/lawrence_wright_2001.pdf+CART+analysis+in+remote+sensing&hl=en&pid=bl&srcid=ADGEESi7tSaxo-AWuKixjmIvUCVT8C4lD-OvkbBMX3fWR4965FgZ897lRt745vN64PKHU_f2Tk7mEP4dc2hVF9Nkh2KyaQn6XUTO3iRY6fs4lh4LCMVQVbLFPTGXmVrXsCyL-Vqmi4ZT&sig=AHIEtbRklGFc7k_DZBxuieiT2W3D8QCs0w https://docs.google.com/viewer?a=v&q=cache:D3oiK_RQl1kJ:www.montana.edu/spowell/pdffiles/lawrence_rse.pdf+CART+analysis+in+remote+sensing&hl=en&pid=bl&srcid=ADGEEShBqrf6dYVI4hrr76LjpZA7FrX3VFbt1mZdqu26bz5Ohl9kZHCA02SYV0mVD8wqOq4n53SLtMRKQpkuA4hDSieLRzWTYc9KPDmSnO17kBAdoRspJPppnJmktB2TBjawtjbRZSgs&sig=AHIEtbR8lerlDVQprH4Mor9LEYQX1XqQ5A I wish that'll help. Try Google if not. About DEM, did you try the tutorials? Quote Link to comment Share on other sites More sharing options...
rahdhitya Posted February 8, 2013 Author Report Share Posted February 8, 2013 Thanks rahman for your help. I had some papers similar with those and i already tried that software a month ago. Lately, i found that eCognition 8.7 had CART algorithm inside, so i dont need to use extra software to do decision tree analysis. But i still failed to do that in eCogintion. Btw, do you have DEM for eCogintion tutorial? I am happy to see Quote Link to comment Share on other sites More sharing options...
rahmansunbeam Posted February 8, 2013 Report Share Posted February 8, 2013 Tutorial home of eCognition. http://community.ecognition.com/home/training-material In this thread, you'll find these lines, 1.) Create the desired image ratios/indexes/principle components/tasseled cap transformations etc for whatever image type you are using. For Landsat analysis at the moment we have about 30 such indexes/ratios defined.2.) Get visual inspection training samples from Google Earth or field survey or wherever... store as shape file... bring in to the project as thematic layer... use "assign class by thematic layer"... use "classified image objects to samples"... use "Feature Space Optimization" and find a reasonable smaller set of the best ratios/indexes. (eg. For Landsat SLAVI, Classical Zabud, MaxDiff etc...) 3.) After multi-segmentation manually set up a CART/SVM algorithm at object level that uses only this optimized set of ratios/indexes features and run it. 4.) Clean-up with the "customized merge by class algorithm" available from the community rule-sets. 5.) Having achieved an initial classification with CART/SVM at object level, re-run at pixel level and look at major differences as this may indicate objects that might require additional segmentation or reclassification. Start playing around with reclassifying the problem areas using other eCognition techniques. quite self-explanatory! More of CART/SVM and other classification algorithm, http://community.ecognition.com/home/classifier-algorithm-cart-svm 1 Quote Link to comment Share on other sites More sharing options...
nichess Posted August 21, 2013 Report Share Posted August 21, 2013 In eCognition the CART Analysis is called "Feature Space Optimization" - you can find it here: Classification >> Nearest Neighbor >> Feature Space Optimization. Create Classes Create Samples Select Features within the Feature Space Optimization window Calculate FSO Apply Features to NN-Feature Space ore to Classes Quote Link to comment Share on other sites More sharing options...
terefice Posted December 16, 2014 Report Share Posted December 16, 2014 please upload imagery tutorial Quote Link to comment Share on other sites More sharing options...
terefice Posted December 26, 2014 Report Share Posted December 26, 2014 In eCognition the CART Analysis is called "Feature Space Optimization" - you can find it here: Classification >> Nearest Neighbor >> Feature Space Optimization. Create Classes Create Samples Select Features within the Feature Space Optimization window Calculate FSO Apply Features to NN-Feature Space ore to Classes please upload imagery tutorial Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.