Abstract
In this paper, expansion of feature vector using Rajski distance image is proposed to improve the accuracy of landcover classification using polarimetric SAR image data. To indicate a reason for the introduction of Rajski distance, simulation of Rajski distance value for the combination of fundamental scattering models is executed, and the results are inspected by comparison with calcurated value from some real polarimetric SAR data. To compare the classification accuracy for the different feature vector ; one of them is constructed only by amplitude images and the other is added Rajski distance images, the availability is shown by Euclid distance method and maximum likelihood method as supervised classifier. The expansion of feature vector proposed in this paper is able to be realized using amplitude images only since this method does not require any phase information.
