Radar Target Range Profile Identification Study

Because HRRP (High-Resolution Range Profile) can provide more accurate target structure and shape information, research on Radar Target Recognition based on Range Profile is of significance for both theory and practice. Discussions on technologies of Range Profile and Radar Target Recognition have been made in this thesis; its content and originality are synthesized as follows:1. Research carried out on Range Profile and Data Processing has been discussed in the thesis.The sensitivity of Range Profile toward the status of the target can have negative effect on Target Recognition. Therefore, it is necessary to investigate the relevant data processing technologies before the recognition of Range Profile. After the investigation of the Range Profile features which is based on the Scattering Center Model, analysis has been made on how Power Transform improves the identification function of Range Profile from the perspective of Signal-to-Noise ratio. Research has been made on techniques of the alignment of Range Profile, and on the application of Autocorrelation Coefficient in the target identification of Range Profile as a novel translational shift invariability feature. Research has also been made on techniques of Dimension Reduction, in particular, on Dimension Reduction of Range Profile which is based on Kernel Canonical Component Analysis.2. Research has been made on Range Profile Classifier.How to identify the targets and increase the recognized rate is the problem that Classifier design should deal with after the acquisition of HRRP features. In this thesis, the descriptions of research on the widely used K-nearest Neighbor Classifier have been made. It is suggested that a Mahalanobis K-nearest Neighbour Classifier can be put into use based on the statistical properties of Range Profile.Descriptions of the experiments on raw data are given in some chapters including Chapter 6, analyses of the experiment results are also provided in this thesis. These experiments have testified the following conclusions:1) Power Transform can improve the identification of Range Profile; 2) When doing Dimension Reduction of the Range Profile features, Kernel Canonical Component Analysis is better than the traditional Canonical Component Analysis in view of its ultimate recognized rate;3) As a method of classification, the Mahalanobis K-nearest Neighbor Classifier has higher recognized rate than the Euclidean K-nearest Neighbor Classifier.Summaries are made and prospects of the Target Identification of Range Profile have also been put forward in the final part of the thesis.