| Polarimetric Multi-Angle Synthetic Aperture Radar(PMA-SAR)has received extensive attention and research in recent years.PMA-SAR uses multiple polarization channels to collect echo data,and employs techniques such as circular,curved or nearlinear trajectories and beam pointing control to achieve Multi-view observation.Compared with single-view single-polarization SAR,PMA-SAR can obtain richer target feature information,which is more conducive to the identification and classification of targets,and therefore has higher research value.Artificial Target Detection(ATD)is an important and widely used in urban planning,resource exploration,disaster monitoring,military reconnaissance and other fields,and is a research hotspot in the field of SAR.As an important step of artificial target detection,the accuracy of feature extraction greatly affects the effect of artificial target detection.Therefore,it is of great theoretical significance and practical value to effectively utilize PMA-SAR multi-view image information for feature extraction and realize efficient artificial target detection.However,the anisotropy of the target information obtained from PMA-SAR multi-views and the poor resolution of the images obtained based on the echo data in the limited view range lead to the inability of the traditional single-view polarized SAR artificial target feature extraction methods to be directly applied to PMA-SAR.In this paper,we address the above problems and challenges,and conduct a relevant study around PMA-SAR feature extraction,and propose two types of multi-view feature-based PMA-SAR feature extraction methods.view angle feature-based PMA-SAR artificial target detection methods,and the main work is as follows:1.The imaging mechanism of PMA-SAR as well as the polarization data characterization method and polarization target decomposition method are introduced,followed by the analysis of multi-view scattering models of typical man-made and natural targets,and an existing PMA-SAR ATD algorithm is introduced.2.In order to effectively utilize the rich information brought by the PMA-SAR multiview images and bring into play the multi-view scattering characteristics of PMA-SAR in artificial target detection,an artificial target detection method based on the difference of multi-view features is proposed.Through polarization channel image fusion and super pixel segmentation classification,the accuracy of multi-view feature extraction is improved and target separability is enhanced;multi-view scattering vector similarity and coefficient of variation are used to quantify inter-view feature differences of targets and combine them with scattering intensity features to construct multi-view target features;finally,the judgment of target type is completed based on multi-view inter-target feature differences.The proposed method can effectively utilize the multi-view features of targets to achieve efficient artificial target detection,and the algorithm has excellent performance in terms of quality factor,accuracy,false alarm rate,missed detection rate,and computing efficiency indexes,as verified by experiments on measured data.3.In order to solve the problem that the inaccurate description of polarization features of traditional PMA-SAR full-aperture images affects the effect of artificial target detection,an artificial target detection algorithm based on fused multi-view image polarization features is proposed.The algorithm successively obtains the multi-view fused image by anisotropic view detection and incoherent accumulation of isotropic views,and then obtains the original polarization features using Polarimetric Target Decomposition(PTD),and then converges the image space by Convolutional Sparse Representation(CSR)to introduce image spatial information into the polarization features to enhance the polarization features,and finally perform the target type judgment based on the enhanced polarization features to achieve the extraction of artificial targets.The quality factor,accuracy,false alarm rate,and missed detection rate of the experimental data show that the proposed algorithm can effectively achieve multi-view fusion,polarization feature acquisition and optimization,and the optimized polarization features can more accurately reflect the real situation of the target and effectively improve the accuracy of artificial target detection compared with the original polarization features. |
