A Fast Generation Algorithm of Radar Images for Ship Recognition

This paper presents a fast algorithm for radar image generation. In this approach, the three-dimensional (3-D) complex scattering function of a complex man-made target is first predicted by a high-frequency EM technique, and then projected onto the two dimensional (2-D) image projection plane (IPP). A fast point-spread operation using 2-D window function is finally made in the frequency domain to obtain a radar image. The developed algorithm is applicable to the online prediction of templates as well as to establish a plentiful training data base for machine learning based automatic target recognition. As an example, a ship recognition approach based on 2-D hot scattering point image is presented by using the proposed algorithm. Experimental results for simulated and TerraSAR-X images demonstrate the usefulness and practicability of the proposed algorithms.