Study On Target Feature Extraction And Recognition Method Of Wideband Polarimetric Surface Movement Radar

The wideband polarmetric technology combines the advantages of the high resolution and polarization measurement,which increases the description accuracy of the targets.Combination of high resolution with polarization has become one of the research focus in RATA community.This dissertation studies the methods of the feature extraction and classification based on the wideband polarmetric radar in depth.The main content is summarized as following five parts:In the first part,the concepts,background and significance of target feature extraction and recognition are set forth.Then the current status of the study and the technical means of feature extraction recognition are expounded.At the end of this part,the main work of this dissertation is introduced.In the second part,several effective wideband features which could be applied in recognition of ground targets are put forward.Firstly,wideband signal processing methods are briefly introduced.Secondly,the wideband features which includes length of the support zone,elevation,the structure of the target scattering centers and the energy ordering are extracted based on the HRRP after the wideband signal processing.Stability and validity of these features are verified by the measured data.And these features can be used to increase the recognition rate of the target.The physical characteristics can be reflected by the features extracted,which can lay good foundation for the following steps of recognition.In the third part,a method of full-polarimetric(FP)information reconstruction using compact polarimetric(CP)data collected by dual-polarimetric(DP)radar system.Methods of extracting CP and DP features which includes scattering entropy,average scattering angle and structural similarity are studied based on FP feature extraction methods.The experimental result shows that reconstruction of CP is feasible and the features extracted are stable and valid which have better performance in classification than DP features.The classification performance can be improved by extracting wideband features reflecting the structures and the scattering characteristics of the targets more accurate.In the forth part,a recognition method of multi-feature and multi-level is proposed.Suitable features are chosen aiming at the recognition method of different levels based on the wideband features and polarization features,which can have better performance in RATR.This recognition method mainly includes two levels.The first level is the multi-feature recognition based on the feature threshold,which includes the judgement of the point targets and distributed targets and the judgements of the targets and interference to recognize the targets.The second level is multi-feature recognition based on the support vector machine(SVM).The experimental result based on the measured data shows that the method could achieve good performance in classification and the feasibility and the validity of the method are verified.The last part presents a conclusion of this dissertation and a description of the future work.