[QUOTE]tarsierspectral wrote Hi, I calculated NDVIs for two images. I want to create 5 classes. They are 1. nonvegetated areas that became vegetated 2. vegetation increased 3. no change 4. vegetation decreased 5. vegetation areas became nonvegetated This is how I am imagining this: nonveg to veg - NDVI values that changed from =0.2 veg increased - NDVI values that changed from >=0.2 to > their previous value No change - NDVI values that are the same for both dates veg decreased - NDVI values that changed from >=0.2 to veg to nonveg - NDVI values that changed >=0.2 to Is this possible in ENVI? Thank you [/QUOTE] ENVI is one of the pioneer remote sensing software, its obviously possible. Open your difference NDVI image and just use interactive density slicing from tools>color mapping> desity slice. More information in help file. http://geol.hu/data/onlin...Density_Slicing.html After applying desired class and range, save the data as classification image. DONE.

[QUOTE]shreshai wrote tarsierspectral wrote Hi, I calculated NDVIs for two images. I want to create 5 classes. They are 1. nonvegetated areas that became vegetated 2. vegetation increased 3. no change 4. vegetation decreased 5. vegetation areas became nonvegetated This is how I am imagining this: nonveg to veg - NDVI values that changed from =0.2 veg increased - NDVI values that changed from >=0.2 to > their previous value No change - NDVI values that are the same for both dates veg decreased - NDVI values that changed from >=0.2 to veg to nonveg - NDVI values that changed >=0.2 to Is this possible in ENVI? Thank you ENVI is one of the pioneer remote sensing software, its obviously possible. Open your difference NDVI image and just use interactive density slicing from tools>color mapping> desity slice. More information in help file. http://geol.hu/data/onlin...Density_Slicing.html After applying desired class and range, save the data as classification image. DONE. [/QUOTE] Hi shreshai, Thank you for that but for some reason I do not have "tools" on the menu bar. I have a toolbox on the right and when I search for "color mapping" or "density slice" it doesn't find such tools. I am using ENVI 5. Do you know where I can find this tool in ENVI 5?

OK, I figured out that density slice in ENVI 5 is called Raster Color Slices. But I can't see any place where I can tell it to look at ndvi of each image and apply classes depending on how they have changed. How do I do that?

If you want to use the Raster Color Slices tool, you would apply it to the difference image that you made using the Change Detection Difference Map tool. If you want to set up classification rules for lots of different types of change, based on the NDVI values in the before and after NDVI images, you might want to think about using the Decision Tree tool instead. Also, I highly recommend checking the online documentation for each of these features, if you are considering using them. - Peg Exelis VIS

[QUOTE]Peg wrote If you want to use the Raster Color Slices tool, you would apply it to the difference image that you made using the Change Detection Difference Map tool. If you want to set up classification rules for lots of different types of change, based on the NDVI values in the before and after NDVI images, you might want to think about using the Decision Tree tool instead. Also, I highly recommend checking the online documentation for each of these features, if you are considering using them. - Peg Exelis VIS [/QUOTE] Thank you Peg. I don't necessarily want to use the Raster Color Slices tool, I want to use any tool that will do what I specified in my first post. Sounds like Decision Tree tool will work better. I will check the documentation. Thank you again

[QUOTE]tarsierspectral wrote Peg wrote If you want to use the Raster Color Slices tool, you would apply it to the difference image that you made using the Change Detection Difference Map tool. If you want to set up classification rules for lots of different types of change, based on the NDVI values in the before and after NDVI images, you might want to think about using the Decision Tree tool instead. Also, I highly recommend checking the online documentation for each of these features, if you are considering using them. - Peg Exelis VIS Thank you Peg. I don't necessarily want to use the Raster Color Slices tool, I want to use any tool that will do what I specified in my first post. Sounds like Decision Tree tool will work better. I will check the documentation. Thank you again [/QUOTE] In my opinion, since you just want to divide difference NDVI image ( which you have already created) into different classes ( threshold values that you have already determined most probably by visual inspection), density slicing your difference NDVI image will suffice. you can start your ENVI interface in old format. From the Start menu, choose All Programs > ENVI 5.0 > Tools > ENVI Classic (or ENVI Classic + IDL). Then just use density slice i as suggested. Remember Keep it stupid simple (KISS) rule, no need to get fancy with DT, thats what i think for this particular task.

[QUOTE]shreshai wrote tarsierspectral wrote Peg wrote If you want to use the Raster Color Slices tool, you would apply it to the difference image that you made using the Change Detection Difference Map tool. If you want to set up classification rules for lots of different types of change, based on the NDVI values in the before and after NDVI images, you might want to think about using the Decision Tree tool instead. Also, I highly recommend checking the online documentation for each of these features, if you are considering using them. - Peg Exelis VIS Thank you Peg. I don't necessarily want to use the Raster Color Slices tool, I want to use any tool that will do what I specified in my first post. Sounds like Decision Tree tool will work better. I will check the documentation. Thank you again In my opinion, since you just want to divide difference NDVI image ( which you have already created) into different classes ( threshold values that you have already determined most probably by visual inspection), density slicing your difference NDVI image will suffice. you can start your ENVI interface in old format. From the Start menu, choose All Programs > ENVI 5.0 > Tools > ENVI Classic (or ENVI Classic + IDL). Then just use density slice i as suggested. Remember Keep it stupid simple (KISS) rule, no need to get fancy with DT, thats what i think for this particular task. [/QUOTE] Can you give me an example how I would determine using density slicer which NDVI values in first NDVI file became higher or equal 0.2 in the second NDVI file? I can't see any place in the density slicer how I would do that. I do want to keep it simple. So if there is a way to do it using density slicer I would like to use that. Thank you

Hello, lets first be clear you have said that you already have difference NDVI image ( one image) out of two NDVI images. The density slider can only set values in single image into different classes depending in intervales ! 0.1-0.4, class 1; 0.4-0.7, class 2, 0.7-1, class 3 so on. It cannot do logical comparision between two files! If you have two NDVI files, then do like this. NDVI values range from -1 to +1 t1 = NDVI time 1 t2 = NDVI time 2 since, negative values are not vegetation make values less than zero as zero.This you can do with band math like (b1 lt 0)*0. do same for t2. Now you have two images where values are positive and most probably represent vegetation. example how I would determine using density slicer which NDVI values in first NDVI file became higher or equal 0.2 in the second NDVI file? This has to done in band math. ((t2-t1) ge 0.2)*1 Above expresion will give you what you want has binary image !

[QUOTE]shreshai wrote Hello, lets first be clear you have said that you already have difference NDVI image ( one image) out of two NDVI images. The density slider can only set values in single image into different classes depending in intervales ! 0.1-0.4, class 1; 0.4-0.7, class 2, 0.7-1, class 3 so on. It cannot do logical comparision between two files! If you have two NDVI files, then do like this. NDVI values range from -1 to +1 t1 = NDVI time 1 t2 = NDVI time 2 since, negative values are not vegetation make values less than zero as zero.This you can do with band math like (b1 lt 0)*0. do same for t2. Now you have two images where values are positive and most probably represent vegetation. example how I would determine using density slicer which NDVI values in first NDVI file became higher or equal 0.2 in the second NDVI file? This has to done in band math. ((t2-t1) ge 0.2)*1 Above expresion will give you what you want has binary image ! [/QUOTE] Hi shreshai, I said that I calculated NDVIs for two images not difference of two NDVIs, although I could do that too, but since the difference map would not allow me to do classifications based on NDVIs from two different times that's why I questioned how I would do that. Can I do this whole classification using band math? nonveg to veg - NDVI values that changed from =0.2 veg increased - NDVI values that changed from >=0.2 to > their previous value No change - NDVI values that are the same for both dates veg decreased - NDVI values that changed from >=0.2 to =0.2 to

Yes it can be done with band math but better to write IDL function if you are familiar with that. With band math it will be very complex but it can be done. First case & third case in one go ((t1 lt 0.2)*t2)gt 0.2)*1+ (t1 eq t2)*3 This will give u image with 0 where either of expression is not valid, 1 where first case is valid, 3 where third case is valid. It should work, though i have not tried it.

[QUOTE]shreshai wrote Yes it can be done with band math but better to write IDL function if you are familiar with that. With band math it will be very complex but it can be done. First case & third case in one go ((t1 lt 0.2)*t2)gt 0.2)*1+ (t1 eq t2)*3 This will give u image with 0 where either of expression is not valid, 1 where first case is valid, 3 where third case is valid. It should work, though i have not tried it. [/QUOTE] OK, so wouldn't the decision tree be easier?

if u think dt is easier, by all means. i rest my case.

[QUOTE]shreshai wrote if u think dt is easier, by all means. i rest my case. [/QUOTE] Just so you know, I appreciate the help you provided anyway. I was able to accomplish this using dt. Once I figured out the syntax it was quite easy actually. Thank you

Here are some methods you can look at: 1) Band Threshold to ROI 2) Band Math 3) Decision Tree Classifier