Multi-to-two-LDA for HRRP radar target recognition

High resolution range profile (HRRP) plays an important role in the field of radar automatic target recognition (RATR). In order to overcome the shortcoming of serious degradation in accuracy when processing a multi-target-recognition problem with LDA, this paper proposes a novel method named multi-to-two-LDA (MTTL), which can transform a problem of multi-target-recognition into a series of two-target recognition problems. The input data is 14 different features extracted from 3 targets' HRRPs. After a multi-target-recognition problem is converted into 3 two-target-recognition problems, LDA processing is performed immediately, and followed corresponding Bayes classifiers are trained. Experimental results show that the proposed method does not only improve the recognition accuracy greatly when dealing with a multi-target-recognition problem, but also keep a high and stable recognition performance under the condition of using few training samples.