Hello Safwan Paramban, how you doing?

I think that one approach to your issue would be to use the trial version of the eCognition, since the software has an user-friendly interface that will let you segment your image without the need of great expertise in the subject. Unfortunately the trial version will last only for 30 days, but that should be plenty of time to accomplish your goal. You can download and get more info about it from here: http://www.ecognition.com/free-trial

And there is also plenty of good video tutorials like: https://www.youtube.com/watch?v=DYtqXUPXVtw

As a open source alternative, you could also use SPRING, which has a poor GUI (Graphic User Interface) but is really robust in its algorithms, and will surely get the job done. It also has plenty of documentation on its download page: http://www.dpi.inpe.br/spring/english/

I've also heard of OTB (Orfeo Toolbox) as a good alternative for classification and segmentation, but never used it. But in case you want to give it at try, here is the link: http://www.orfeo-toolbox.org/otb/

About the urban feature extraction from the scene, I would suggest you to use the (R) band, wich is pointed out by the literature as a good way of extracting synthetic and impervious features from the scene.

If you want more in-depth info on the subject, I would also suggest reading the work of https://www.researchgate.net/profile/Qihao_Weng

Best regards,

David

Dear David

Many thanks and  I really appreciated your contribution.

I am doing fine.

I downloaded the Ecognition trial version and applied the segmentation,ya its getting good result,but problem is have to export the segmented output to get the pixel value for the connecting the Spectral Mixture Analysis.

Unfortunately in  trial version export tool not working.

And you mentioned the SPRING Open source,I am trying in that software.

Again i face few problem,Probably you can give few more valuable  suggestion ,please

>As i mentioned my image have 1 meter resolution so when the time of Impima module using output image parameter sampling value how much i have to give ?

>When the we doing the Segmentation the Similarity value and Area(Pixels) how to decide?

I am trying to do using this Link...................

http://staff.wwu.edu/~antonon/envr442_/ENVI/442_segmentation_ENVI_acme2.htm

Once Again..Many thanks to you dear.

Regards

SAFWAN

>As i mentioned my image have 1 meter resolution so when the time of Impima module using output image parameter sampling value how much i have to give ?

I guess you can set the sampling value to 1, that may give you an output image with the same 1m spacial resolution of the input image, as of stated by the tutorial:

"In the Output Image Parameters section, click on all 8 bands and change the Sampling to 1. This will retain the same 2m by 2m pixel size in the output image."

>When the we doing the Segmentation the Similarity value and Area(Pixels) how to decide?

The image used in this tutorial isn't from a urban area, but I guess you can use the same parameters as him:

"In Region Growing, set the Similarity box to 10 and the Area(pixels): box to 25. This Areas(pixels) sets the smallest region (polygon) that will be delineated in the image."

From my experience, this is a sort of empirical part of the process. You will probably need to repeat this step of the segmentation until the parameters that you have set to the region growth algorithm give you suitable segments.

You can find more about how this segmentation algorithm works on this video: https://www.youtube.com/watch?v=VaR21S8ewCQ That may help you decide on which parameters you should use.

Dear David

Many many thanks to you for your valuable suggestions.

I tried as per the Journal's different parameter value they are giving.

But when the Spring Image has importing in to Map window of the Spring Software I am getting the Image has compress type.

Yaa we have to repeat same steps of the segmentation untill the parameters that  i have set to the region growth algorithm give me suitable segments.

Sure Let me try....and I will let you know.

Thanks a lot.

SAFWAN