Research On Techniques Of Target Detection For SAR Image Of Harbor Region

Harbor region belongs to the typical environment with land and sea mixed, which is very complex. It is now a challenging subject of radar area to detect and distinguish ships of interest in harbor region. According to the application requirements for airborne or missile-borne radar to detect ships in harbor region, the SAR image of harbor region is firstly simulated, based on which, a succession of tasks including sea-land segmentation, region of interest (ROI) extraction, precise coastline extraction and ship detection are analyzed. The main work includes the following aspects:(1) The background and significance of the research are mainly introduced in chapter 1, and a review of existing techniques of SAR image simulation, SAR image edge detection, and SAR image ship detection is outlined.(2) The simulation of SAR image of harbor region is studied in chapter 2. According to the prior information acquired from the gray-scale features of SAR image of harbor region, a series of backscattering coefficient models of land, sea, jetty, dock and ships are firstly defined. Then an echo of harbor region is created based on target scattering model, and SAR image of harbor region is acquired by processing the echo with the conventional SAR imaging algorithm. Based on the statistical distribution of clutter, land and sea clutter in typical environment are generated and added into land and sea respectively. Accordingly, SAR image of harbor region is finally obtained under various Signal-to-Clutter Ratios (SCR). Studies on the simulation of SAR image of harbor region have provided adequate experimental data for the subsequent research.(3) Sea-land segmentation for SAR image of harbor region is studied in chapter 3. Based on the SAR image of harbor region obtained, methods of coastline detection and threshold segmentation combined with mathematical morphology are firstly adopted in order to realize the sea-land segmentation. Then the harbor ROI is extracted based on the geometry features of jetty and dock. Regarding the miss-alarm problem incurred by the sea-land segmentation when ships are located closely to the land, a method of extending strip area is adopted to realize the accurate extraction of coastline, and meanwhile the miss-alarm problem is also avoided.(4) Ship detection for SAR image of harbor region is studied in chapter 4. The algorithms of Constant False Alarm Ratio (CFAR) based on clutter statistic model are firstly introduced. Based on the framework for ship detection of SAR image of harbor region, a method of K-distribution-based local adaptive CFAR detection algorithm is adopted, which solves the false-alarm problem incurred by SAR speckle noise and sea spike, and a comparison of detection performance under different SCRs is conducted by simulation experiments.(5) A conclusion of the thesis and a description of the future work are presented in the last chapter.