Abstract: Dynamic multi-object recognition based on machine vision is a very important research direction in the field of pattern recognition. In this paper, the multi-object identification of robotic fishes in water is studied and A HSV color space model is proposed. By re-modeling the original color space, the main light interference in the image can be effectively removed, and the shadow of the robotic fish can be removed, making the background segmentation more accurate. The improved algorithm has stronger robustness than the original matching tracking algorithm

Keywords: Robotic fish, target recognition, machine vision, HSV, deep learning