Image recognition is playing a more and more important role in all kinds of detection methods, and it is also a key link in artificial intelligence. It provides decision-making basis for computer or mechanical equipment to take correct action. In aquaculture and food processing industry, image recognition has the advantages of directviewing, real-time and adaptability, which provides powerful guarantee for the development of intelligent agriculture and food safety. This article mainly uses the image recognition, take Red Sea Bream as an example, carries on the damage detection to the fish surface, provides the reliable basis for the further processing. In this paper, YOLO v3 algorithm is used to improve the speed and accuracy of fish surface detection, thus providing a new idea for the development of intelligent aquaculture processing. Compared with the traditional way of food processing industry, the efficiency and accuracy of using image recognition to judge whether the surface of fish is damaged, whether the whole fish is intact and whether the whole fish meets the requirements are greatly improved. Finally, we need to solve the next problem, and the direction of development put forward their own ideas.