In addition to the above mentioned methods, you can use ArcGIS analysis tools (overlay) after converting the raster data to vector.

Hi, one approach is to simply difference the two rasters assuming they are labeled as discrete class values (e.g. 1 to 4 for four class types), then compute number of pixels where there is a non-zero difference. Then for each pixel you can determine whether it changed classes, to what class it changed too, and then build up many different quantitative / descriptive measures to describe the total change between the images. For example, you can compute percentage of pixels for a given class that changed to another class--> e.g. 30% of class 1 changed to class 2 or 20% of class 2 changed to class 1... or 80% of all pixels at time 1 changed to a different class at time 2, etc;...or you can report the change in areal coverage between the two images; E.g. at time 1, 1300 pixels at a resolution of 1 m were labeled class 1 covering 1300 m^2 and at time 2 only 900 pixels (900 m^2) were labeled class 1 resulting in a reduction of (400 m^2) in land cover for that class. etc...

Thanks for your reply, i got your point, kindly tell me what tool to use for this difference?

If you work with ENVI, first of all you must save your maps in ENVI format. Then in "basic tools" you can find the change detection tool which let you achieve the change detection map and the statistics (how many pixels changed, considering pairs of classes). If you want to evaluate the accuracy of the change detection map you need a ground truth, that is a sample of known changed pixels for each possible transition between classes. In this way you can create a confusion matrix, as for a classification map.

Hi, I'm not sure how to explicity do it in ENVI / ERDAS. I looked briefly online and saw this whitepaper which states that in ERDAS IMAGINEyou can use the matrix operation from GIS analysis menu to compute change between two thematic maps; this link is to white paper:

http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=4&ved=0CEEQFjAD&url=http%3A%2F%2Fwww.erdas.com%2FLibraries%2FWhite_Papers%2FChange_Detection.sflb.ashx&ei=OkCHT7aIO8eN8AHr55i2CA&usg=AFQjCNGjuXtymeyOvTmutCzLKpf1oPD-Cg&sig2=uSoDFk34a0Gfk8cVvF8Hcw

Also, if you are familiar with open-source GRASS GIS you can import the two rasters into it and then use its map algebra toolset to compute the difference and other statistics; "r.mapcalc" is the function name --> see this link for more details

http://grass.fbk.eu/gdp/html_grass64/r.mapcalc.html

you can also access it via the GRASS gui. Here is link to GRASS software for download

http://grass.fbk.eu/

Or if you are familiar with MATLAB and have access to it then you can import in your two classified (labeled) rasters as matrix and compute difference and other statistics.

Hi Jennifer

The change may be expressed as difference or ratio of two rasters. To quantitatively express the difference you may change the raster to vector and compute the difference in area. Else you may use modeler in ERDAS to carry out the same operation. Hope this may help you. Dr Ghosh

There is a video in youtube describe the way how to do this operation. Tape detection change with Erdas imagine or Eradas imagine only and enter. Ok

The simpliest way is to do image difference , you can use any classified image to do so ( supervised , unsupervised ) , hope this will help

This is old one but still relevant paper about change detection methods: http://www.tandfonline.com/doi/abs/10.1080/01431168908903939

There is an interesting change detection approach that is calculating lanscape metrics after classification and compare changes in metrics.

In addition to the above mentioned methods, you can use ArcGIS analysis tools (overlay) after converting the raster data to vector.

@ Hadjkouider Mehamed i already had seen that video, but could not understand due to German language unfamiliarity.

Sorry, there is an other in English. And you can see the steps and understand without listen to the language

Hi Jennifer

Find the difference in area between two time points using raster area calculation ( open your classified image- goto Raster menu-attributes-edit-add area column). your raster area will be calculated based on your units ( recommended UTM for units to be meters). if your are not convinced with this ,please do specify your problem in detail.

@Jennifer

Hi jennifer

I want to tel you that I have the the video ( in my labtop) from youtube but it's big and i can't sed it by e-mail .

can you tel me wat can I do to shard with video with you. realy there are a good details in it about your quastion.

Ok, but how can I do it please .......???

@ Hadjkouider Mehamed thanks. so kind of you. please tell me if i tell you my email id, can the transfer of file be possible? or else tell me your linkedin id, and add me there, and share video with me.. thanks.

realy, I don't know because the file is so big it's about 170 MO in 30' long. In few days, I try to give you web-link if I found it OK. Thanks’

@Jennifer Stanley.

Hi, her is it the adrasse ''http://www.youtube.com/watch?v=vrIAyMHG7pk'' I don't Know if you have got befor or not.

I think it's important for youe topic .

Thanks'

In Arc, go to Spatial Analyst Tools-Zonal-Tabulate Area. You then specify the two input layers (supervised classification from two dates). You will get a transition matrix for all possible combinations.

Hopefully, you recognize that a shortcoming of this "classify and compare" approach is that the accuracy of your change detection is the product of the accuracy of your two supervised classifications. So, for example, if the accuracy of the classification for each date is 80%, the accuracy of your change detection is 0.8 x 0.8 = 0.64.

Alternatives that have a higher accuracy include image differencing (e.g. Cohen et al. 2002. Ecosystems 5:122-137) or trajectory-based change detection (Kennedy et al. 2007. Remote Sensing of Environment 110:370-386)

I hope that this helps.

Go to Change detection analysis of ERDAS Imagine software, there in attribute it will show the classwise change properly.

@Jennifer Stanley.

Hi, I sent to you the adress, is that what you looking for ?

yes thanks alot. itz very helpful... God bless you.

Ok, Thanks' for your prayer ...thank you .

In Envi EX ... in the toolbox double click thematic change , the select file pannel apears...browse....classification image time1...open file.....spatial subset (samples and rows), classification image time2....apply change detection....Only Include Areas That Have Changed....

Idrisi Selva and TerrSet are very useful for this task. Good luck.