Extraction Of Structural Features By Rotating Bar Mask From A Satellite Image

In this paper we present a new feature extraction method in a spatial domain which has a very good ability in detecting thin linear structures (roads and rivers), as well as rectangular shapes (agriculture fields 1. This new spatial feature extraction method is implemented by applying a rotating bar mask with varying both horizontal and vertical sizes to a segmented TM image (labeled image), which is obtained by using unsupervised cluster seeking algorithm in the principal component feature space.