Behavior Tracking Of Large Yellow Croaker Based On Machine Vision

The ocean contains extremely rich species resources.The ocean is by far the largest ecosystem.My country’s ocean area is about 4.7 million square kilometers,of which fishery resources play an important role in marine resources.However,with the influence of human activities,fishery resources have shown a declining trend in recent years.The identification and tracking of fish has broad application prospects in the fields of intelligent feeding of aquaculture,research and development of fishery resources,detection of fish behavior,and upgrading of aquaculture industry structure.This paper mainly studies the following aspects:(1)Build an experimental breeding platform for large yellow croaker,collect data of large yellow croaker under static water flow,and conduct identification and tracking research on large yellow croaker.A fish identification and tracking algorithm based on the Py Torch framework is adopted.Two network frameworks,Res Net-50 and YOLOv5,were used to detect the behavior of large yellow croaker in the experimental breeding environment.The quality of the model is judged according to the target confidence,tracking effect,and detection time.The accuracy of the algorithm is shown by indicators such as Precision,Recall,and m AP in the confusion matrix.The advantages and disadvantages of the two models are analyzed to provide a basis for optimizing the network structure.(2)For the behavior of the large yellow croaker in the actual breeding process,the data of the large yellow croaker in the breeding cages will be collected in time and at a fixed point.After screening,preprocessing is performed on the 2 680 image data taken.Using the Py Torch framework,the YOLOv5 algorithm optimized for network parameters and structure is used.Through field tests,the accuracy rate is over 95%,and the detection speed is 61 FPS.(3)Use PyQt5 to design a GUI visualization interface,call the Pyinstaller script,and package the program files.Through the file selection and prediction function buttons in the interface,the large yellow croaker is detected.The test results show that the tracking and recognition effect is consistent with the prediction.At the same time,the large yellow croaker quantity distribution map is generated,which can be used for data analysis.Using the optimized YOLOv5 algorithm S model to do real-time tracking and detection of large yellow croaker,the effect of distinguishing overlapping fish bodies and tracking and matching is obvious.Not only the detection accuracy of the original model has been improved,but the detection efficiency is also greatly accelerated.Through the optimized YOLOv5 algorithm,real-time and accurate monitoring of the actual number distribution and aggregation location of large yellow croaker in cages is performed,and relevant detection information is output.By analyzing and researching it,the behavior,posture and living habits of large yellow croaker in culture can be obtained,which will help to improve the healthy breeding of large yellow croaker and promote the development of intelligent feeding.