Research On Target Tracking Method Of UUV Underwater Recovery Based On Vision

Due to the limited size of the Unmanned Underwater Vehicle(UUV),the energy carried by itself is limited,so it needs timely refueling.Usually the UUV refuels after recovery,thus the method of recovery is one of the key research directions of UUV.How to track the recovery device is the main point in the UUV recovery process.This thesis is based on a new UUV independent recovery method,folk-carrying-pole recovery,to achieve target tracking during the recovery process.The specific research contents are as follows:The monocular vision system of folk-carrying-pole recovery of UUV.Method of positioning according to the L-shaped lamp shelf is used to achieve the folk-carrying-pole recovery of UUV.The L-shaped lamp shelf and Tritech's underwater camera Tornado formed a monocular vision system,whose target light source is D550 underwater light flashlight fixed on the L-shaped lamp shelf.On the basis of prepare work,the model of monocular vision system is introduced and a positioning method of the monocular vision system is raised.Besides,a new algorithm of measuring the heading angle is put forward in the condition of the presence of a trim angle,and the simulation verifies the rationality and effectiveness of the improved heading angle algorithm.Target light source recognition during the recovery process.Due to the complexity of the underwater environment,the images acquired by the camera need to be pre-processed first,including image filtering and image segmentation.Since the hydroacoustic noise is mostly impulse noise,median filtering is used to filter the image.Taking the size of the light source in the actual application into consideration,usually they vary in size,the segmentation operation is performed by threshold segmentation to solve this problem.A target recognition method based on slope and relative distance is proposed,which can recognize the target sources and calculate their pixel coordinates.On this basis,two pseudo light source removal methods based on geometric relationship and light source area,depending on the differences of the location of the pseudo light source,are designed respectively.The target recognition method of convolution filtering is proposed on the condition that the target light source is in the background light environment.Target tracking method during recovery.First introduced the basic theory of kernel correlation filtering,then a target tracking method based on kernel correlation filtering is designed.This method is mainly divided into four steps,training samples,training kernel correlation filter,quick detection of target light source and update the model of the kernel correlation filter.In order to adapt to the scale transformation of target light source during camera movement,a KCF algorithm based on target scale transformation is designed.Another KCF algorithm based on background subtraction is raised to reduce the influence of the target light source being occluded.Target tracking simulation test in recovering.Firstly,the system hardware composition and overall process of the target tracking test are introduced.Secondly,the UUV target tracking test scheme is designed.The scheme is roughly divided into three parts,detect the viewable area of underwater camera,UUV target tracking in single degree of freedom and UUV target tracking in the case of target occlusion.Then target tracking of UUV simulation tests under different conditions including longitudinal and lateral change,vertical change,heading change,occlusion change,and pseudo-light source state.Finally,the rationality and effectiveness of the KCF algorithms are verified by analyzing the test results of the pool.