Research On Detection Of Single Scattering Part In SAR Data

Synthetic Aperture Radar(SAR) can operate under all weather and all day which has become an important tool for battlefield surveillance. Though the technology of SAR automatic target recognition(ATR) has made great progress, it cannot meet the requirement of the elaborate battlefield environment. To recognize the SAR target quickly and precisely and make judgments on the possible changes on its parts has important intelligence value in the real battles. Thus, the realization of more precise and high level SAR ATR has important meaning.Model-based SAR ATR has become the hotspot in the field of SAR ATR for its good performance under extended operating condition(EOC). The construction of the target's part-level 3-D electromagnetic scattering model can help to realize the part-level SAR ATR. The model describes the electromagnetic scattering property of the whole target and the scattering parts of the target under special frequency and pose which makes it possible to detect the single scattering part of the target. This thesis focuses on the methods of the detection of single scattering part for part-level SAR ATR. Following are the main works of this thesisFirstly, the construction of the 3D electromagnetic scattering models of target and its parameterized expression are described. Then the frame for the detection of single scattering part is designed based on the model's properties which depicts how to detect the single scattering part from the 2-D image or features predicted by the model.Secondly, the method of the detection of single scattering part based on matched filter is proposed. Matched filter is a classical algorithm in signal detection. The predicted image of single scattering part by the model is employed to construct the matched filter. The filter slides in the 3-D parameter domain to enhance the precision of the detection. CLEAN algorithm is employed to deprive the interference of other scattering parts to the present detecting one. Then, the method is applied to the simple tank model and its electromagnetic simulated data, dark-room data and MSTAR interfering data to detect single scattering part. The results demonstrate that the proposed method can well detect the existing scattering parts in the measured image and find the absent ones.Lastly, the method of the detection of single scattering part based on feature matching is proposed. The feature of the single scattering part is constructed according to the its image predicted by the model and a proper rule to evaluate the similarity of the predicted features and extracted features is designed to detect the single scattering part. Then, the method is applied to the simple tank model and its electromagnetic simulated data, dark-room data and MSTAR ?interfering? data to detect single scattering part. The results demonstrate that the proposed method can well detect the existing scattering parts in the measured image and find the absent ones.