A Study Of Feature Extraction And Applications Using Inverse Synthetic Aperture Radar Images

Inverse synthetic aperture radar(ISAR) is an imaging radar using microwave receiving data for the sake of high-resolution images for both space and aerial targets, such as satellite, space station, airplane, ship, etc. It can image, detect, and recognize targets in far-distance, all-time, and all-weather conditions. With the popularization of this technique, its applications expand from a unilateral imaging to in-depth surveys of target’s components, motion parameter estimation, and so on. So far, it is valuable to study interesting components and their related geometric structures, scattering features, motion states for the theories and practice of analyzing effective loads, components classification, etc. In this thesis, we utilize two-dimensional ISAR data and images as digital resources to focus on features of targets, parameters estimation, components fusion and detection in order to analyze targets’ physical sizes, status, and local structures. In what follows is the outline of this thesis:Chapter 1 introduces the research background and purpose of this thesis. It concerns the synopsis of the development of ISAR imaging, the post-processing of ISAR images, and the works of our projects in detail.Chapter 2 proposes the applications of feature extraction on cross-range scaling and phase autofocusing. These applications are based on the reconstruction of cost function of target’s motion, followed by an evaluation to obtain highly accurate rotational angle between two sequential sub-aperture ISAR images. Upon these processes, we are ready to acquire effective cross-range resolution to conduct cross-range scaling and then estimate real size. Meanwhile, estimation of motion parameters by this method makes it available to improve the autofocusing performance through cancellation of phase items.Chapter 3 proposes the processes of different sub-aperture data in short coherent processing intervals(CPIs), e.g., analysis of target’s structures, geometric changes, motion parameter estimation. These results are applied into fusion of target’s structures under long CPIs for a relatively complete ISAR image containing all possible structures to enhance the recognition accuracy and analyticity. We first estimate the motionparameters and geometric distortion factor to acquire modified ISAR images after corrections of geometry and aspects. Then we adopt effective fusion to deal with sequential ISAR images with different imaged structures and energy levels to obtain an optimal ISAR image with all possible structures and a highly focusing quality.Chapter 4 proposes a novel method for interesting components detection and recognition based on ISAR images. The first step is to conduct local registration with a feature framework by integral feature information and local feature information of template components. As the feature framework shifts, we calculate the corresponding probabilistic accuracy each time. Next, we evaluate the final detecting accuracy by a defined criterion. This survey attempts to deal with component detection of space targets due to their contrasty structures and relatively high resolutions of ISAR images.