U can use Erdas 2011, it detects segmented regions automatically

Hiii

Dear Md Siddqur Rahman,

Many Many Thanks for your contribution.

I had tried to do the Segmentation in Erdas Imagine 2014.

But I want to detect the Shape of the Urban Feature from the satellite imagery.

Please find Segmented result....I have attached.

From this how i can do the Detect the Shape of the feature..

Please..Let me know.....

Thank you dear

Dear Safwan

You may use segmented features for a supervised classification.

What do you think about using auto detect/ grow at regions for segmented features. I think ur utimate aim is to 

Dear Mr Siddiqur Rahman

Myne ultimate aim is to detect the Shape of the Urban feature from 2005 and 2012 satellite image and do change detection focused with Shape.

And I have to connect the this output image in to Spectral Mixture Analysis Concept.

Means I have to estimate the proportion of each pixel covered in each polygon(mainly Buildings) of the area.

Thank you Dear

Hi Safwan,

I think you will have to start with moderate expectations here. ENVI is quite limited when it comes to object-based image analysis (as is Erdas Objective). ENVI tries to find a segmentation that fits the entire image, which is not realistic – you naturally have objects of different sizes in most images (certainly in urban scenes). So far I have not seen good results with ENVI. Erdas Objective is also not very useful. It’s essentially pixel-based throughout, and only in the end pixel clusters are converted into objects of some sort. If you want real OBIA ability you’d have to go to eCognition, even though it’s quite expensive (maybe try the free trial version to see what can be done).

Good luck, Norman

To gain information about a suitable size of objects in image segmentation, you can refer to the following publication. In this publication, I introduced different thresholds and investigated the results of all cases.    

Article Automatic building extraction in dense urban areas through G...