Research On Sea Target Detection And Recognition In High Resolution SAR Images

SAR has been widely used in many fields,such as military reconnaissance and marine applications,because of its advantages of all-weather and all-time active operation.Target detection and recognition has also become a research hotspot.In this paper,we take the ship target as the research core,carrying out the work of sea-land segmentation,ship detection and ship size classification and recognition.Firstly,we review the current research status of ship detection and recognition,compare and analyze the advantages and problems of existing methods,and summary the characteristics of SAR image and SAR surface target.Then,we analyze the factors affecting ship target detection and recognition in detail,such as environment,SAR system and ship itself.After that,the basic process of ship target detection and recognition is introduced.There are a large number of strong scattering objects and buildings in the land area of sea surface SAR image,which is not conducive to ship target detection.Based on the difference of gray scale and edge density between ocean and land,this paper uses the method of sea-land segmentation based on edge detection and region growth to remove the land part,so as to avoid interference to the subsequent detection.Traditional detection is greatly affected by clutter,and often loses part of the ship shape information.Moreover,the detection result often has some false alarm.To solve this problem,we propose a ship detection algorithm based on super-pixel segmentation and random forest clustering.In this algorithm,the super-pixel segmentation can ensure that the ship has a complete and accurate shape,and at the same time,raise the ship detection algorithm from the traditional single pixel to the super-pixel level,which reduces the computational complexity.The purpose of random forest clustering is to quickly find rare super-pixel blocks of the ship.The experimental results show that the proposed algorithm has good anti-noise performance,high detection rate,and can well retain the shape characteristics of ship targets.After obtaining the ship detection results,the shape features are extracted to classify the size of the ship.