Knowledge-Based Radar Target Classification

In this chapter, knowledge-based algorithms are developed for the problem of target classification for ground surveillance Doppler radars. Two sources of knowledge are presented and incorporated within the classification algorithms: (1) statistical knowledge on radar target echo features and (2) physical knowledge, represented via the locomotion models for different targets. An algorithm that combines both sources of knowledge is proposed. Various methods to incorporate these sources of knowledge are presented. Maximum-likelihood (ML) and majority-voting decision schemes were applied for target classification. The proposed classification approaches were tested using real data of radar echo recorded by ground surveillance radar, which include various target classes, such as walking person(s), tracked or wheeled vehicles, animals, and clutter. The combined approach, which implements both statistical and physical prior knowledge, provides the best classification performance, and it achieves a classification rate of 89%. The majority-voting decision scheme, which is suboptimal, is found to perform better than the ML criterion due to modeling mismatch.