The Research Of Ship Detection And Recognition Algorithm In SAR Image And Its Hardware Implementation

Synthetic aperture radar has the advantage of all-day and all-weatherr abservaation,and is widely used in Fmilitary investigation.With the rapid development of SAR imaging technology,it has beem able to obtain a large number of high-resolution SAR images,but the interprteiatior work such as image dertecting,locating and feature extracting can not meet the needs of the current SAR image procestsing.so the study of SAR interpretation is of great practical significance.This subject relics on the pre-research project of National Defense ××× to carry out the research on ship detection and necognition in SAR images.This paper mainly solves the problems exising in ship detection,location and feature extraction,aiming at the situations of sca reconnaissance of the autonomous missile-borne platform,the wide variety of marine warship target and the complex target characteristics.This dissertation focuses on two aspects:Firstly,we studies the target detection,location and feature extraction algorithms involved in the automatic detection and identification system of SAR image ships.Secondly,we studies the hardware implementartion of the above algorithm.The part of algorithm implementation:Firstly,this dissertartion studied the ship detection technology of CFAR(comstant fase alarm rate)in the SAR image,and analyzes the problem that the adaptive threshold obtained by using a single barckground clutter distribution model eads to a large number of false alarm targets in CFAR detection results.In this paper,the geeometric features of ship are integrated into the CFAR detection algorithm,and the improved algorithm can reduce the false alarm target under the condition of guaranteeing the detection rate.Secondly,aiming at the target location of the ship,we studied several regional connected labeling algorithms.The two-time scan labeling algorithm needs to deal with a large muber of equivalent temporary labels which is a complex processing and may consumes a lot of time.This dissertation based on the two-time scans,and adopted recursive method which just need one-time scan to extract the region information.From the extracted target area information,the number of ship targets in a SAR image and the information of each target ship boundary can be clearly counted.Lastly,for the characteristics of the target ship,we studied the geometric characteristics of the SAR image,the gradation characteristics and texture feature,finally extracted texture statistics of the ship by GLCM(Gray level Co-occurrence Matrix).The part of hardware implementation for the algorithm,because in the large-scale SAR scene,the ship's target detection,location,and fearure extraction algorithms have to calculate a large amount of data,so the software's inherent serial processing method makes its operating speed unable to meet the needs of practical applications.To avoid this situation,based on the in-depth study of the above algorithm,this paper firstly uses FPGA parallel processing to port the improved algorithm to Alteral's FPGA development board,whose chip model is Stratix IV and series model is EP4SGX530HH35C2,then we designed a SAR image ship detection and feature extraction hardware system.This hardware system includes three components:a CFAR target detection unit,a target area information extraction unit,and a target feature extraction unit.Secondly,after the entire project was synthesized through the Quartus II software of Alteral Corporation,the results showed that the maximum clock frequency of the entire system can reach 116 MHz.Finally,the designed system was verified by using the simulation verification platform,and verification results conform with the design requirements.