Research On SAR Target Detection And Recognition Based On Feature Extraction Of Fractional Time Frequency Spectrum

SAR Automatic Target Recognition(ATR) is the key component of SAR Image translate and analysis. Because of the different imaging theory, SAR image is polluted by coherent spot noise, geometric distortion and sensitive to azimuth which leads to the difference between background scattering intensity and target scattering intensity. So it is significance to find a way extracting effective information of target and make a resonable translate of SAR image. This thesis mainly research on the feature extraction of optimal fractional time frequency domain and the method of SAR target detection and recognition.On the basis of pioneers’ relevant scientific research. The thesis combines Time Frequency Analysis(TFA) with Fractional Fourier Transform(Fr FT), to increase the flexibility of this transient signal analysis method. The method is been used for SAR target detection, and make an in depth research on this area. The mainly research work and contribution as follows:(1) Research on the basic theory of SAR ATR. In order to understand the characteristics of the SAR images, we learn the imaging mechanism, the principle of noise production and the noise denoising method. Also do some analysis on the characteristics of SAR target, build a theory foundation for designing a SAR ATR system.(2) Research on the TFA method and the principle of Fr FT, also an in depth research on Fractional Time Frequency Analysis(Fr TFA). On the basis of the Gabor Transform which has the lowest computation complexibility, we propose the Fractional Gabor Transform(Fr GT) and successfully promote it to 2-Dimensional image processing.(3) Further research of optimal implemention method of Fr GT is been proceeded. In view of window function designing and optimal transform order searching, a lot of research work is been finished. At last, we unfold an experimental simulation and data analysis.(4)To reduce the dimension of time frequency spectrum and make it visualization. After the further research of the optimal implemention, the spectrum’s feature extraction method is been explored. We choose the optimal feature and guaranteed the detection precision. Propose the Optimal Fractional Gabor Transform(OFr GT), At last, we unfold an experimental simulation and data analysis.(5)Carefully research on the designing method of classification module. Discuss the combination of characteristics and decision fusion. At last, we unfold an experimental simulation and data analysis.The test results of real SAR image library indicates that the proposed detection and recognition method is available and Fr TFA gives a new way on SAR image processing.