Research On Feature Extraction And Identification Of High Resolution Radar Target

Radar for target recognition is a main development trend of modern radar system with high resolution. At present, the target recognition based on high resolution radar in military and civil aspects have a certain degree of application. This paper uses the measured high resolution radar echo data experiment, focus on the target feature extraction in radar recognition system and fusion, and the use of support vector machine classifier for the selected feature. The main work includes the following contents:First of all, on the basis of research on the formation and properties of radar target range profile of radar target one-dimensional distance, found the like contains a large number of target structure and shape information, therefore, for the one-dimensional distance image for feature extraction. In the study of one dimensional range profile based on the shift sensitivity, using the translation invariant feature extraction method to overcome this problem. Extracting the target power spectrum, center distance feature and the amplitude spectrum difference feature, and the study found that the three kinds of feature extraction of targets with differences.Secondly, in order to obtain the effective characteristic, this paper presents an improved principal component analysis method based on rough set fusion applied in radar target identification. Feature selection is the most important in radar target feature recognition system is the link target. The selected features should not only be able to describe the target, but also have some differences with other similar targets. In does not affect the characteristics of information content at the same time also should try to reduce the feature dimension for using the least feature to contain the most effective information, and then do a fast, efficient and accurate target recognition. Power on the target spectrum, the center distance and amplitude spectrum characteristics of principal component analysis is divided into feature difference, then the theory of rough set attribute reduction of target features based on making the fused features with a large number of target information, and greatly reduces the dimension of feature space, thus ensuring the superiority of the fusion feature.Finally, detailed introduces the principle and application of support vector machine classifier, and the application of three different algorithms with a variety of kernel function of one-dimensional range profile of target fusion characteristics and single feature are extracted, the research results show that the improved rough set feature fusion of principal component analysis in identification of not only better than other properties. Based on theSyndrome, and the dimension is greatly reduced. This not only improves the recognition rate of the system, but also saves the storage space of the recognition system, reduces system complexity degree.