An Object Oriented Approach for Multi-Channel and Multi-Polarization NASA/JPL POLSAR Image Classification

摘要

This paper presents an object oriented approachrn(OOA) for classification of multi-channel and multi-polarizrnation NASA/JPL POLSAR images. Some test results inrnTaiwan are also given and analyzed. It is concluded thatrnthis approach can utilize as more information of both lornw- and high-levels involved in all images as possible forrnimage classification and thus provides a better classificatiornn accuracy. For instance, the OOA has a better overall clarnssification accuracy(98.27%) than the nearest-neighbor clasrnsifier(91.31%) and minimum-distance classifier(80.52%).