Ship Detection In SAR Image And Its Design And Implementation Using DSP

With the development of aerospace technology,the remote sensing data are increasing rapidly.The traditional off-line methods at the ground stations can't meet the real-time processing requirement in extracting intelligence information from remote sensing data.Therefore,this paper mainly focuses on the on-orbit remote sensing data real-time processing in SAR ship target detection and its implementation by using DSP.The main work of this paper is as follows.(1)Considering the efficiency,computational complexity,resource consumption in DSP design,various SAR ship target detection algorithms are compared and analyzed.The global CFAR detection algorithm based on a k-distributed clutter model is selected as the pre-screening algorithm.To dismiss the false alarms,a sequential feature identification method is adopted to effectively remove false alarms,which is closely related to the performance of target segmentation.Thus,this paper introduces an image segmentation algorithm based on change detection technology,and makes improvements in the process of generating the change detection measure.(2)Bow discrimination is a follow-up study for ship detection,which is of great significance to ship classification,speed estimation and target association.Since the bow of a ship is sharp while the stern is blunt,and their contours are both parabola-like curves,this paper uses the quadratic function to fit the curves based on a least square error method.The coefficient of quadratic term can be used to identify the bow of the ship effectively.In order to obtain an accurate ship target contour,the GVF snake deformation algorithm under the elliptic constraint is adopted to extract the SAR ship contour.(3)In view of the performance,resource and efficiency requirements of on-orbit SAR ship detection,this paper selects TMS320C6678 DSP as the embedded hardware platform.Firstly,SAR ship detection codes realized on PC platform are simplified,rewritten and transplanted to the DSP platform.Then,considering the DSP memory characteristics and access priority relations,memory access optimization is explored to improve the speed of detection.Finally,the detection speed is further improved by codelevel compilation and instruction scheduling optimization.