A shape-based object recognition for remote sensing
Joji Iisaka, Takako Sakurai-Amano · 2002
A robust method to describe any shape which is observed in remote sensing images has been developed. In this method, an object is expanded in a series of pattern primitives or structural elements of different sizes. The first few expansion coefficients of a small number of different primitive expansion sets reasonably approximate the shape of an object in remotely sensed data in terms of shape differences. The method has been applied to a transformed JERS-1 SAR image.