Study On Ship Detection Technology Based On Region Segmentation For SAR Images

Because SAR has the ability to monitor the marine environment all day,all-weather,it is very important for the economy,people's livelihood and national defense to carry out the research of ship detection technology based on SAR images.Since the birth of SAR technology,scholars and research institutions at home and abroad have carried out a lot of research on ship detection technology in SAR images,and made a series of important achievements.However,due to the blurring of the target contour in SAR images,the traditional method is still difficult to achieve good detection results.In addition,with the development and progress of SAR technology,the SAR image data acquired by people is growing exponentially,and the image data is developing in the direction of large space and high resolution.The traditional pixel based processing method is more and more incompetent for the future development trend.In order to overcome the bottleneck of the existing methods,this paper studies the ship detection technology based on the SAR image region segmentation method,and makes a detailed and in-depth exploration of the superpixel method and the level set method,based on which new ship detection methods of SAR images are proposed.The main research work and achievements are as follows:The existing methods for ship detection in SAR images are systematically introduced.This paper systematically reviews the development of SAR technology,introduces the existing methods of ship detection in SAR images,and analyzes the main problems of the existing methods.Aiming at these problems,combined with the future development trend,the main research directions and contents of this paper are introduced in detail,and the main innovations of this paper are given.The superpixel generating method for SAR images based on pixel saliency difference and spatial distance is produced.Starting from the existing superpixel generation algorithms,this paper analyzes the characteristics of SAR images and the shortcomings of existing methods,and proposes a superpixel generation method for SAR images based on pixel saliency difference and space distance.In order to solve the problem that it is difficult for the existing superpixel methods to generate accurate superpixel segmentation in the non-uniform edge and texture regions,this paper proposes a Gauss kernel weighted local contrast measure operator,which can greatly enhance the dynamic range of the pixels in the non-uniform regions and at the same time effectively suppress the speckle,and provide favorable conditions for the accurate detection of the fuzzy edges of the target.In addition,an adaptive local compactness parameter is proposed to solve the problem of poor effect of superpixel methods in the mixed region.The experimental results show that compared with the existing methods,our method can achieve more accurate superpixel segmentation under different noise environment at almost the same operation efficiency.The sea clutter statistical analysis of SAR images based on superpixel is studied.Based on the traditional method of sea clutter statistical analysis of SAR images,this paper innovatively proposes to carry out sea clutter statistical analysis of SAR images based on superpixel.In this paper,more than 80 kinds of distribution functions in the powerful Sci Py scientific calculation database are fitted with the superpixel sample data of sea clutters.By comparing the fitting errors,the most suitable distribution model of the superpixel level sea clutter is obtained,namely,the Johnsonsu distribution.Through multiple sets of data fitting experiments,the fitting ability of the Johnsonsu distribution of sea clutter data is verified.The method of Constant False-Alarm Rate(CFAR)ship detection based on superpixel is studied.Based on the superpixel method and Johnson Su distribution model,this paper studies the superpixel level CFAR detection algorithm.Different from the existing CFAR detection methods based on superpixel,this paper adopts a new ring topology structure,background superpixel uses Johnsonsu distribution function modeling,and carries out comparative experiments with the traditional K-CFAR and the state-of-the-art SP-CFAR algorithm,the results show that our method can achieve higher detection rate and lower false alarm rate.The method of ship detection based on superpixel level Markov Random Field(MRF)is studied.The traditional MRF method is based on the pixel space relationship,which is difficult to achieve accurate segmentation.In this paper,a superpixel level MRF ship detection algorithm is proposed,in which the superpixel is used as the minimum analysis unit to solve the problem that the traditional MRF method can not use the high-level image information.In addition,a hybrid nonlocal field distribution function is proposed to make the segmentation result more accurate.Compared with the traditional MRF method,the experimental results show that the superpixel level MRF method has higher detection accuracy than the traditional MRF method.In terms of inshore ship detection,our method can achieve higher efficienc than traditional MRF method.Inshore ship detection based on the level set method and visual saliency for SAR images is studied in detail.In order to further improve the accuracy of the inshore ship detection in SAR images,this paper proposes a detection method based on the level set method and visual saliency.Starting from the local binary fitting model and the improved local contrast measurement algorithm,the main steps of the algorithm include:firstly,using the principle of downsampling to initialize the SAR image quickly;secondly,using the level set method based on our proposed visual saliency fitting energy to segment the SAR image;thirdly,extracting candidate targets through adaptive threshold;finally,obtaining the final detection results through area screening.Experimental results show that this method can achieve higher detection rate than traditional K-CFAR and the state-of-the-art Multi Parameter Weighted Image Entropy(MVWIE)algorithm.By comparing with our superpixel level MRF method,experimental results of inshore ship detection show that our proposed level set method can obtain more accurate contour results when the ship target integrity is poor.