Font Size: a A A

The Study On Target Identification For Low Resolution Radar

Posted on:2011-01-13Degree:MasterType:Thesis
Country:ChinaCandidate:X P ZhaoFull Text:PDF
GTID:2178330332460735Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
Radar target recognition is always a research focus in the radar detection filed. The high resolution radar echo contains abundant target information, which can achieve precise target classification, while the low resolution radar echo is lack of resolution capability in vertical and horizontal, so, the precise target details such as shape and structure cannot be obtained. A large number of experiments implemented both at home and abroad have been proved that it is feasible to obtain rough radar target classification through low resolution radar. Therefore, the main purpose of this paper is to achieve radar target classification according to the objective size based on the low resolution radar.For the problem of low resolution radar target recognition, this paper starts with the radar echo forming principle to explain the reason why the low resolution radar echo contains limited target information, and also expounds the feasibility of rough radar target classification. Through analyzing lots of actual radar measurement echoes, this paper mainly studies the feature extraction methods based on one-dimension radar echo, and some feature extraction methods based on two-dimension profile image are also presented. The recognition results show that the proposed feature extract methods can provide the effective support for the design of radar target recognition system.Before extracting the one-dimension echo feature, some abnormal echoes are eliminated firstly which always emerge during gathering process, and then the target echo features are extracted mainly from three different aspects:echo shape characteristic, encoding characteristic and wavelet decomposition characteristic. Firstly, three different kinds of target echoes are analyzed separately based on width, shoulder width and area, and we put forward with the necessity of the echo group characteristics as the final feature vector. Secondly, we adopt the Kth order differential encoding method and the improved encoding method based on freeman, respectively, to extract some features such as entropy and LZ complexity. Thirdly, wavelet decomposition is proposed to extract features for the radar echo is a typical non-stationary signal. The energy is computed for wavelet decomposition coefficients at different scales, and the energy of different frequency signals is normalized as feature eigenvector. In addition, p-norms waveform-entropy of the echo signal is extracted for wavelet decomposition as characteristic. Before extracting the two-dimension profile image characteristics, an automatic pretreatment algorithm is proposed, and then three different features are extracted, such as simple shape depicted, gray image information and moment invariant.Finally, in view of the complexity and variability of radar target characteristics, BP neural network is adopted as the classifier. The actual measured radar echoes are recognized as three different categories based on one-dimension radar echo characteristics and two-dimension profile image characteristics. According to the recognition results obtained by the different characteristics, this paper analyzes the advantages and disadvantages of the various characteristics, which has the certain significance for the design of radar target recognition system.
Keywords/Search Tags:Low Resolution Radar, Feature Extraction, Wavelet Decomposition, Target Recognition
PDF Full Text Request
Related items