Three-dimensional Target Reconstruction Based On ISRA Image Sequence

Inverse synthetic aperture radar(ISAR) is a signal processing technique for moving target in range and Doppler domain imaging. ISAR is used for scenarios in which the radar is fixed and the targets, e.g., ships, aircrafts, satellites, space stations, tanks are in motion. Because the ISAR imaging is not affected by climate, time and distance constraints, its application prospect has widely attracted researchers’ interests. The 3D reconstruction techniques will be an important research branch of the ISAR application field which could greatly enrich the classification methods and improve the recognition performance of targets.This dissertation mainly discusses the ISAR 3D reconstruction of image sequences of the motion objects. The ISAR 3D reconstruction based on image sequences is indeed a process of reconstructing the scatting point in the target. Considering the properties of the actual motion process and the characteristics of ISAR imaging, ISAR 3D reconstruction scheme is proposed in this dissertation. According to the imaging principle of the single base ISAR, the generation process of ISAR 2D image sequences is elaborated. The observation matrix model for scattering points which are stable and traceable on the target is established. Then the necessary conditions of the 3D reconstruction from the 2D image sequences are given. In the presence of the speckle noise in the process of ISAR imaging, the traditional Canny edge detector performs badly. Comparing to the Canny edge detector, the Gauss Gamma shaped bi-windows detector shows more robustness to the speckle noise. After Gauss Gamma shaped bi-windows filtering, a gradient based technique is applied to obtain corner information(i.e. the projection point of the target feature points in the image plane). The Lucas-Kanade optical flow scheme which is based on image Pyramid tracks the target feature points in the image projection plane. This scheme solves the imprecise local positioning problem which is caused by the large motions of scattering points in the projection image plane, and it relaxes the conditions of the motion space of the adjacent image frames. After tracking the target feature points, the estimated projection coordinates of feature points form the observation matrix. The factorization method decomposes the observation matrix to the three-dimensional structure based matrix and the motion matrix. Through excluding the non-unique decomposition and the non-positive definite problems, the factorization method obtains a 3D structure matrix which could be used to recover the solid scattering model. Simulations based on a hotel and space station 2D image sequences demonstrate the effectiveness of the proposed 3D ISAR image reconstruction scheme.