Underwater Target Detection And Tracking Technology Research Based On Machine Vision

The target detection and tracking technology is a hot topic of computer vision, and theunderwater target detection and tracking is an important content of underwater robots，which not only is widely used in underwater robot navigation、underwater video monitoringand many other fields, but only plays an important on researching and salvaging target inthe pool of nuclear power plant. In the support of the Hi-tech Research and DevelopmentProgram of China “plant multi-function underwater climbing robot”, the underwater targetdetection and tracking technology is studied in this paper.In this paper, the main work and innovative results are as follows:（1）According to the principle of underwater optical imaging and other variousfactors, this paper not only analyzes the characteristics of the underwater imaging, but alsoresearches the image clearer algorithm based on underwater imaging model whichestimates and wipes out the influence of the scattered light and light changes on theunderwater imaging by wavelet transform.（2）According to the complex underwater radiation environment, this paperresearches the adaptive window underwater target tracking algorithm based on color. Onthe basis of traditional color matching tracking algorithm, the algorithm not only finds thebest position by color localization algorithm and tracking accurate algorithm to improve thetracking accurate, but also calculates the width of window according to the trackingwindow size change algorithm which basically realizes the size of tracking windowchanges with the size of target in the image.（3）According to the characteristics of underwater imaging blur and unstable, thispaper researches the underwater image segmentation algorithm based on two-dimensionalfuzzy Otsu. It extends the Otsu from one dimensional to two-dimensional and combineswith the fuzzy theory, which combines double advantage of the Otsu and the fuzzy theory.And then according to the geometric characteristics of the underwater target, this paper getsthe auxiliary information of the gripper grabbing target which include the position of masscenter、 the rotation angle of wrist and the rotation angle of motor that controls the gripperopening and closing.（4） Many experiments which include underwater target tracking、 imagesegmentation and feature extraction are performed using LabVIEW software platform, butalso the feasibility and effectiveness of algorithms is verified.