Visual Tracking Technology For Underwater Fish Monitoring

With the improvement of people's living standard and the great leap of technology,more and more attention has been paid to the intelligent aquaculture monitoring and life habits of underwater fish.At present,underwater fish monitoring is mainly through manual monitoring.Divers' tracking and shooting,many underwater environmental risk factors are high,and divers' tracking may disturb the fish,which can not reflect the most real life rules and health of fish and other important data.Therefore,the intelligent underwater fish tracking robot and its core visual tracking algorithm are constantly explored.In this paper,the technology of machine vision recognition and location for underwater vehicle fishing is studied.The main contents are as follows.During the pre-research period,a new underwater swimming fish tracking algorithm based on kernel correlation filtering algorithm was proposed.The main idea of the algorithm is to track the two ends of the swimming fish separately,which not only effectively improves the tracking efficiency,but also greatly improves the position accuracy and scale accuracy.The image processing speed of the algorithm can reach 29 frames per second,the average accuracy can reach 70.01%,and the target tracking success rate can reach 87.11%.In the process of further learning,the algorithm based on correlation filtering is improved.The main idea is to optimize the scale factor to 7,which will improve the speed of the algorithm calculation,and basically will not affect the tracking accuracy.The image processing speed of the algorithm can reach 23 frames per second,the average accuracy can reach 66.71%,and the target tracking success rate can reach 91.83%.In summary,the two tracking algorithms proposed in this paper can be applied to the automatic tracking of fish,both satisfy the requirements of tracking accuracy and tracking success rate,and can track the moving fish target in complex environment.The improved tracking algorithm based on kernel correlation filtering algorithm can recognize and locate the target fish faster,but based on correlation filtering.The improved wave tracking algorithm has a higher success rate in identifying and locating target fish.It can meet the fast feedback requirements of the control system,and has high stability,high success rate and good robustness.