Research On Small Target Detection Algorithm Based On Sonar Image

Due to the increasingly open international situation,the competition among countries is becoming increasingly fierce.Because of its importance,marine resources and ports have always been the focus of international disputes,which makes the underwater security of national offshore areas particularly prominent.With the continuous miniaturization of underwater weapon equipment,small targets such as frogman and underwater robot have become important means of underwater attack in the "asymmetric" strategy.Therefore,every country takes the detection of underwater frogman as the research focus.One of the main means of underwater detection is sonar technology.The sonar system is used to image the underwater detection environment,and the sonar image is used to detect the small target of frogman.It is helpful to build the underwater monitoring system of the port in the coastal area,and also provides the corresponding technical guarantee for the national underwater safety.Therefore,in this thesis,the sonar data is displayed as an image,and the new algorithm of small target detection in sonar image is studied.The main contents of this thesis are as follows:Firstly,it is studied in this thesis that the coordinate transformation algorithm of the improved nearest neighbor interpolation.Considering that the field of view of sonar system is a sector area in polar coordinate system,the improved nearest neighbor method is used to transform the data matrix of sonar image formed by beamforming,which eliminate the image hole caused by direct conversion.Meanwhile it improves the data loss caused by the broadening of acoustic diffusion angle,and the image quality is enhanced,so as to it can help to observe the seabed environment and target.Secondly,an adaptive K Singular Value Decomposition(K-SVD)sonar image denoising algorithm based on residual image is studied.According to the multiplicative noise characteristic of sonar image reverberation noise obeying Rayleigh distribution,sparse representation theory is applied to sonar image denoising algorithm.The experimental results show that the denoised image of this algorithm retains more edge details while removing noise,and it has better evaluation results in subjective and objective indicators.Then,an adaptive threshold sonar image small target detection algorithm based on integral graph is studied.Considering that the sonar image corresponds the echo data to the pixel value,and the echo information of the target is generally stronger than the echo intensity of the underwater background environment,the integral graph isintroduced to estimate the echo and background of the target,calculate the adaptive threshold,and the threshold is used in segmentation algorithm to detect the small sonar target.Experimental results show that the algorithm can effectively detect small targets in the underwater environment,and the use of integral graph can speed up the calculation speed,so as to improve the real-time performance of the algorithm.The detection position of the continuous frame image is connected to get the motion track result of the small target.Finally,the detection algorithm of small target in sonar image based on contrast is proposed.The integrated image of sonar image is used to calculate the pixel value of brightness,and the calculated contrast image is segmented by threshold value to detect small targets.The experimental results verify the effectiveness of the algorithm that improve the detection accuracy of small targets,improve the detection accuracy and achieve better detection results.Aiming at the problem of small target detection in sonar image,this thesis focuses on two new small target detection algorithms.The experimental results verify that these algorithms are effective,so they can be used to construct the underwater monitoring system.