Fish Target Detection and Speed Estimation Method based on Computer Vision

Taking the monitoring of fish status in aquaculture cages as the research background, the survival status of fish is assessed through the estimation of fish swimming speed, and the Yolov3 target detection algorithm and Deep-sort multi-target tracking algorithm based on deep learning are studied to achieve fish target detection. And trajectory tracking, the imaging model of the underwater binocular camera is established, the three-dimensional coordinates of the underwater fish target image point are obtained, and the speed estimation of the fish target is realized on the basis of the continuous tracking of the fish target. Compared with the other algorithms of fish body separation, the research results show that the method proposed in this paper has an average detection accuracy of 96.16% for underwater fish targets, and the average relative error of fish movement speed estimation under natural lighting conditions is 1.60%.