Research On Technique Of Ultra-wideband SAR Shallow Buried Targets Feature Acquisition

Ultra-wide band Synthetic Aperture Radar(UWB SAR), which has the advantages of wide detection area and strong penetrating capability, makes it possible to quickly detect shallow buried targets over large areas from a safe standoff distance. Feature acquisition is the key of obtaining good detection performance of UWB SAR shallow buried targets. In this thesis, technique of UWB SAR shallow buried targets feature acquisition is studied based on the real data achieved by the airship mounted ultra-wide band SAR system.Firstly, the efficient feature of UWB SAR shallow buried targets is analyzed. The theory of efficient feature is proposed by studying both the target and the sensor. The electromagnetic scattering model of shallow buried target is constructed by electromagnetic simulation and the real data. Based on the electromagnetic scattering model, the time domain double-hump feature, the frequency domain sharp-dip feature and the aspect angle invariance feature of metallic landmine is analyzed. Then the space of UWB SAR shallow buried target efficient feature is established. The relationship between the UWB SAR observation model and the space of UWB SAR shallow buried target efficient feature, as well as the relationship between the detection procedure and the space of UWB SAR shallow buried target efficient feature, are discussed respectively.Secondly, the technique of prescreening feature acquisition of UWB SAR shallow buried targets is studied. To solve the problem of the inefficiency of contrast feature acquisition in traditional CFAR, a quick contrast feature acquisition algorithm based on integral image is proposed. After analyzing the disadvantages that it is apt to be interfered by the complicated environment when using the contrast feature in UWB SAR targets prescreening, a UWB SAR prescreening method based on the local structure feature is proposed. The double-hump structures of metallic landmine can not be focused simultaneity, which will make the local structure feature acquisition lacking precision. To solve this problem, a metallic landmine local structure feature acquisition method based on the double-hump feature enhanced operation is proposed.Thirdly, the technique of discrimination feature extraction of UWB SAR shallow buried targets is studied. A frame of multidimensional scattering function estimation based on the time-frequency analyse is proposed. The restriction of uncertainty principle in multidimensional scattering function estimation via the classical time-frequency analyse method is discussed. A UWB SAR targets discrimination feature extraction method based on empirical mode decomposition(EMD) is proposed. The 1D EMD is developed to 2D EMD, and a separate ensemble EMD(SEEMD) method is proposed. The procedure of scattering feature extraction of SAR target scattering centers is presented. A UWB SAR targets discrimination feature extraction method based on sparse representation is proposed. The time-frequency analyse method based on sparse representation is developed to the 2D space-wavenumber domain, and the scattering characteristics of stationary scattering center is sparse represented. Aimed at migration of the near peak and the far peak of the metallic landmine double-hump with the variety of the aspect angle, a sparse representation method of migratory scattering center is proposed.Finally, aimed at that the feature vector dimension is too high when taking the multidimensional scattering function as discrimination feature vector directly, the discrimination feature dimensionality reduction technique of UWB SAR shallow buried targets is studied. The necessity of feature dimensionality reduction is analyzed, and the feasibility of UWB SAR shallow buried targets discrimination feature dimensionality reduction is demonstrated by the intrinsic dimension. To solve the problem that the linear dimensionality reduction method can not process the nonlinear discrimination feature data of shallow buried targets, a discrimination feature dimensionality reduction method of UWB SAR shallow buried targets based on manifold learning is proposed.