There are numerous ways in which an image can be segmented into separate regions and have these regions classified. One approach is a per-field classification where an unknown region with defined boundaries is classified. An alternative approach, and the one used herein, is a pixel-by-pixel classification method in which both segmentation and identification are done at the same time. This allows for a region with an unknown boundary to be classified. This thesis involves the development of a software program using the Grey Level Co-occurrence Matrix (GLCM) to both segment and classify regions in MSTAR (Man-portable and Surveillance Target Acquisition Radar) imagery. The technical background of the GLCM along with a user's manual for the software is included. The GLCM has several parameters that can be varied to affect the classification results that it will produce. These parameters are pixel separation, pixel orientation, number of grey scale values in the image being analyzed, number of pixel combinations used, and the choices of which measurements to use. Results show that the GLCM can be used to both segment and classify an image at the same time, with difficulty occurring at regional boundaries and images containing both rural and suburban areas.There are numerous ways in which an image can be segmented into separate regions and have these regions classified. One approach is a per-field classification where an unknown region with defined boundaries is classified.
|Title||:||Using the Grey-level Co-occurrence Matrix to Segment and Classify Radar Imagery|
|Author||:||Jeremiah R. Ferguson|
|Publisher||:||ProQuest - 2007|