Geometrical definition and computation methods of line target features

Abstract

Line target recognition in discrete space is a difficulty in the field of image processing. Line features in the scene include tracks, roads, rivers, bridges and field boundaries. Usually, target features include amplitude, histogram and statistics characterizations. However, these features are insufficient to describe a line target for a practical image. Utilizing the geometry of target, we present the definitions and computation methods of length, orientation angle, smoothness degree, roughness degree, and barycenter for line target in discrete space. Computational examples show that these new definitions can exactly character line target such that it can provide new theoretical and technical supports for image processing of line target in discrete space.Index Terms-feature extraction, line feature, image processing, synthetic aperture radar (SAR) image.