The Target Feature Extraction And Recognition For High Range Recognition Radar Based On The Kernel Method

The target recognition for radar is the one of the important direction of modem radartechnique, and is the important part of the arms system in the future in the world. The highrange resolution profile obtained from the high range recognition radar contains the moreinformation of target's features. The feature extraction is the key step of the target recognition.It's important of choosing what kind of the feature which can character the target. This paper isinvestigated the applying of the KPCA and KDDA methods to the HRRP. Do experiments withfour targets in the end.In chapter one, the research background and the meaning of the paper is introduced, also itshows us the content of pattern discriminant, the development and the status of radar targetrecognition and the manifold methods of feature extraction.The obtaining of the HRRP and its characters, is introduced in chapter two. Then thestrongly sensitivity in pose and moving of target is analyzed. Lastly the paper summarized themethods of the feature extraction and classification using HRRP.The target feature extraction and recognition for high range recognition radar based on theKPCA method investigated in chapter 3. Firstly, do pretreatment for HRRP, and low the SNR ofHRRP whose feature to be extracted. Then it discusses the theory of the KPCA algorithm. Thispaper has constructed a co-kernel method which can extract the mathematics feature both fromwhole part and the local part. Then the method of the target's feature extraction and recognitionof HRRP radar based on the KPCA is brought forward. At last, do experiment on four targets. Ithas made out that the method could advance the recognition performance.The target feature extraction and recognition for high range recognition radar based on theKDDA method is investigated in chapter 4. The non-linear problem in the input space convertsthe linear problem in the high dimension space by the kernel function. So we can solve theproblem using LDA which we familiar. The KDDA method effectively solves the SSS problemin the high dimension. Besides, the best discriminant vector has the most information we want.Chapter 5 summarizes the whole paper, and points the problem which needs to beimproved.