Fish feeding behavior identification method based on YOLOv5

Abstract

A fish ingestion behavior identification method based on YOLOv5 comprises the following steps that videos before and after feeding of a certain single fish school are collected through a camera, key frame pictures are extracted and marked, and a data set is established; a picture is input into a YOLOv5 network model, after being preprocessed, the picture sequentially enters a Backbone network and a Neck network for feature fusion, then the picture is input into a Head output end, the position, category and confidence of a prediction frame are obtained, weighted non-maximum suppression is adopted to screen the prediction frame, weight parameters are adjusted by calculating a loss function, model parameters are continuously optimized through a verification set, and optimal weight data are obtained; and the test set pictures are input into the trained YOLOv5 network model to obtain a target detection result. According to the method, the accuracy of target detection is high, the YOLOv5 network model is adopted, rapid identification, real-time tracking and prediction can be achieved, actions can be taken in time, and the purpose of improving the breeding benefits is achieved.

Description

technical field [0001] The present invention relates to the technical field of target detection, in particular to a fish feeding behavior recognition method based on YOLOv5. Background technique [0002] In aquaculture, the amount of bait input is an important factor related to the aquaculture economy. Insufficient bait input makes it difficult for farmed fish to grow rapidly. Excessive bait input causes waste and increases costs on the one hand, and deteriorates the environment of the aquaculture water body on the other hand. , It is not conducive to the health and production of fish. The ideal state is to feed on demand. The amount of bait added can just ensure that each fish can absorb a sufficient amount of food without leaving any surplus. However, the ideal state is difficult to achieve, and it is difficult to control only by experience. When the cost of bait increases, even manual feeding is used instead of automatic feeding equipment. Different feeding effects will ...

AI Technical Summary

Problems solved by technology

For visual technology, it is necessary to extract the characteristic parameters of the ingestion pictures and establish an ingestion prediction model. At present, the commonly used models include BP neural network model, support vector machine (SVM) model, etc., but the accuracy is often low, the model size is large, and the running speed is slow. , affecting its application

Images