Research and application of fishery resource assessment based on particle filter algorithm physical habitat model

Abstract

In order to solve the problem of accurately and robustly recovering fish trajectories from videos, the author proposes a physical habitat assessment model for fishery resources based on particle filter algorithm. The idea of “group tracking” was proposed, which divides all targets into several groups based on their mutual distance and processes different groups independently and in parallel, avoiding the extremely high dimensionality caused by estimating all zebrafish parameters simultaneously. The parameters of all individuals in the same group are estimated together, and it handles occlusion by modeling the interaction and occlusion of objects in the same group. The experimental results show that from frame 16 to frame 21, two fish suddenly turn around, especially the fish marked in red; From frame 34 to frame 42, these two fish interact for the first time, but do not intersect; From frame 49 to frame 58, these two fish interact and pass through each other for the second time. The algorithm proposed by the author, GCT, and contrast algorithm CT, can effectively handle the protruding and directional motion of fish. Conclusion: The algorithm is accurate and reliable.