Research on fish visual recognition technology based on YOLOv5 fishery robot

When fishery robots perform the task of fish recognition, they face blurred vision and color distortion caused by underwater impurities, which seriously restricts the visual recognition of underwater fish. To address this issue, a combination of image preprocessing and YOLOv5 technology is proposed to enhance the rapid recognition and adaptability of fishery robots to targets. Firstly,the fishery robot preprocesses the underwater images through a differential channel gain based on the improved multi-scale Retinex algorithm, and then uses the target detection algorithm of YOLOv5 for fish recognition. The results show that the proposed method can achieve accurate detection and clear visualization of fish targets.