RGB (color additive process) combination of temporal SAR data, it is one of the method for identifying change detection.

Hi Mohomed,

There are a lot of ways to detect changes with SAR imagery. Using intensity/amplitudes is one way (ACD - Amplitude Change Detection). You may consider Coherence Change Detection (CCD) techniques, as well . Depends on the application and your data.

Doesn't matter too much, if your data are inensity in dB, amplitude, ect. You may try as well to stack your temporal data and do principal component rotation (PCR).  As Iyyappan mentioned RGB/HSV combination is a way to visually highlight changes.

Tobias

Some links:

http://www.iaeng.org/publication/WCE2013/WCE2013_pp2231-2237.pdf

http://elib.dlr.de/61982/1/schmitt.pdf

http://elib.dlr.de/66399/1/schmitt_fullpaper.pdf

http://www.isprs.org/proceedings/XXXVIII/1_4_7-W5/paper/Scheuchl-185.pdf

Nice paper on the visualization of changes with optical data:

http://www.mdpi.com/2072-4292/6/11/11533/htm

Hi Mohomed,

Lot of methods are there for detecting changes in SAR images.  Even i am doing research in that area only. But am not getting what is the need of converting intensity to db...If u have given that information then it would be feasible to answer.

Regards,

Sumaiya.

Radar Back scatter is best represented by RCS, which is given in DB. Intensity operations are simply like working with grayscale images. However, the target interaction is completely different in a case of a Radar and an optical image. So it depends upon the nature of the targets in your scene what kind of methodology you would require. 

Mostly with SAR amplitude images, you should look for some statistical methods. I am not sure but PCA would not be that significant in this respect because the images would not be coherent on a multi-temporal scale. And changing the component to principal vector will not be appropriate since you would be doing pre-processing I presume.  

Depending upon the nature of the scene, your best option would be D-InSAR or Differential SAR tomography, i.e. if you have those heaps of data. Else I would suggest go for some statistical method. 

Hi Mohomed,

I have used this method for environmental change detection and published a couple of papers in Archaeological Prospection that you can access from my researchgate page. I found the use of dB very effective to correlate changes in backscatter with surface processes, vegetation and soil moisture variations.

Hope it helps.

I would like to add a couple of comments to the already good suggestions and comments made by others. 

I have investigated the calibration and use of remotely sensed data sets since the early 1970s (yes, I go WAY back).  Data sets have included optical satellite and aerial images (visible, nir, and swir spectral bands), thermal, radar (SLAR and SAR --- SLAR is the pre-SAR radar imaging systems), acoustics (side-scan sonar) of ocean and lake bottoms, and other data sets.  As with many applications, including this one, change detection is an important aspect of the use of remote sensing.   I have found that depending on the application CALIBRATION (such as converting raw DNs to dBs) may or may not be needed.  If the user wants to assign a 'physical property' to the pixels (such as surface spectral or backscatter reflectance) then calibration is important; however, to strictly just detect change the important thing is that the DN values of the multi-temporal images be the same.  This can be dBs but can also be the raw DN values as long as the sensor characteristic have not changed.  You may need to do relative calibration and match the date 1 image to a date 2 image and then do change detection.  In my view a change can be detected regardless of what units the image DNs are in; again, they just have to both be in the same units.

Within my profile in research gate you will find several papers I have published on change detection using remotely sensed image pairs.

Pat 

Hi, we also developed some techniques for change detection (both using amplitude and phase). Not just by comparing pairs of images, but also by analyzing time series. We are still working on the subject, with many new ideas. A few papers on change detection can be found here http://www.sarproz.com/scientific-publications/ By the way, published ideas are also implemented in the software you find on that page. If anybody would like to try them in his own cases, please let me know.

Thank you all your suggestions were very useful, I performed the radiometric calibration which provides the db value and simply followed the image difference by applying a suitable threshold which gave me a significant result. Now  i' m working on possibility of the insar coherence change detection compared to ACD, D.perissin i would like to try the method you have suggested.