基于深度学习的鱼类识别与检测的算法研究

Research on fish recognition and detection algorithm based on deep Learning

鱼类分类识别在渔业资源研究、鱼类知识的科学推广、水产养殖加工、稀有物种保护等领域具有广泛的应用前景。针对大菱鲆、黄鳍鲷、金钱鱼、鲻鱼这四种鱼类,利用PyTorch框架为基础,通过ResNet50网络模型,用不同的算法对其进行分类识别,不断对模型进行优化,对四种鱼类训练学习,通过测试其准确率达到96%以上。同时用PyQt5开发了GUI可视化界面,通过界面图片的选择和预测功能按钮的操作,测试结果实际类别与预测类别一致,用DSOD框架做了水下目标实时跟踪检测,提高了对小目标的检测率,同时保持了模型的检测速度,检测结果达到期望。

Fish classification and identification has broad application prospects in the fields of fisheries resource research,scientific promotion of fish knowledge,aquaculture processing,and protection of rare species.In this paper,for the four species of turbot,yellowfin snapper,golden fish,mullet,based on the PyTorch framework,the ResNet50 network model is used to classify and identify them with different algorithms,and the model is continuously optimized.The accuracy rate of fish training is 96%.At the same time,the GUI visual interface was developed with PyQt5.Through the selection of interface pictures and the operation of the prediction function button,the actual category of the test results is consistent with the predicted category.The real-time tracking and detection of underwater targets was done using the DSOD framework,which improved the detection rate of small targets.At the same time,the detection speed of the model is maintained,and the detection results meet expectations.