Study On Automatic Target Recognition Technology Based On HRRP

Radar Automatic Target Recognition plays an important role in the modern warfare environment. High Resolution of Range Profile(HRRP) obtained by broadband radar reflects the target scattering center distribution along the radar line of sight direction, and contains abundant target structure and shape information. Therefore, target recognition based on HRRP has attracted widespread attention at home and abroad.HRRP target scattering characteristics is a starting point for the study of HRRP target recognition. First of all, a brief analysis of radar echo characteristics is discussed when the scatter center model has no change. Then according to HRRP echo characteristics, a method is presented which is a HRRP target recognition algorithm based on the feature combination and SVM(Support Vector Machine, SVM). The incoherent average HRRP feature in frequency domain and central moments feature are extracted respectively in the algorithm. And they are serially combined into new features. SVM is used to perform classification at last. Experimental results with Moving and Stationary Target Acquisition and Recognition(MSTAR) SAR(Synthetic Aperture Radar) HRRP data sets show that the proposed algorithm can improve correct recognition rate in the case of not using the target azimuth, and is an effective method for HRRP target recognition.In addition, the sparse representation is applied to HRRP target recognition. A new sparse representation namely supervised sparse preserving projection(S2PP) based on feature combination is shown in the paper. Because sparse preserving projection(SPP) is a type of unsupervised method, which cannot make full use of the label information especially dealing with supervision problems. Therefore, S2 PP is introduced by using label information on the basis of SPP, which means that firstly, the sparse coefficient is obtained by the linear representation between the other training sample from the same class. Then, the sparse coefficient is introduced to the feature extraction. The feature vectors in lower dimension is achieved through the relationship of sparse reconstruction at last. S2 PP can not only make the extracted lower dimensional features keep the sparse reconstruction character, but also can eliminate the influence of the diffident kinds of targets on sparse representation. Because of the good discrimination ability of feature combination, S2 PP is performed based on feature combination. The experimental results with MSTAR show the effectiveness of the proposed method.