Radar automatic target recognition based on high resolution range profile (HRRP) has been widely studied by researchers at home and broad. The prerequisite of target recognition is a good knowledge of the properties of target data. However, such work has rarely been involved. As we all know, HRRP has the following sensitizations: azimuth, shift and amplitude, and how deals with these three sensitizations problems becomes the difficulty and the hotspot in radar auto-target recognition. This paper puts forward PCA-subspace method which constructs templates according to azimuth, solve the shift problem with the shifting-correlation method and perform amplitude normalization to HRRP. Principle Component Analysis has been involved in this method so as to abstract the main features which can efficiently represent the corresponding target, meanwhile, to decrease the dimensions and get rid of noises. Through a great deal of experiments on real radar data, it has been proved that PCA-subspace method can work efficiently and increase recognition rate, compared with the traditional nearest method.
|