Research On Target Tracking Algorithm Based On YOLOv4 Detection Network

Target tracking is both a key research topic and a research problem in the field of computer vision technology.It is widely used in many fields,such as military,traffic intelligence,security and so on.Due to its outstanding feature modeling capabilities,deep learning is considered an important supporting technology in the field of video target tracking.According to the basic test database(VOT,OTB,etc.)test platform,the accuracy of the video target tracking algorithm based on deep learning is basically ranked in the top.However,there are some problems with this algorithm: a large-scale deep neural network has a great impact on the real-time performance of video target tracking;some neural networks cannot completely segment the target,extract the target features reasonably and accurately identify the target.In response to the above-mentioned related issues,this paper will carry out corresponding research on the target tracking algorithm based on the deep learning neural network framework.The main work includes:First of all,a multi-target tracking algorithm based on YOLOv4 and SORT algorithm is proposed.First,YOLOv4 is used for target detection,and the detected pedestrian data is passed to the SORT algorithm to achieve multi-target tracking.SORT algorithm N can achieve good performance at high frame rates.Through comparison,it is concluded that the multi-target tracking algorithm proposed in this paper can improve the corresponding m AP and FPS.Secondly,it is proposed to use the combination of YOLOv4-MOT and Deep SORT algorithm for multi-target tracking.It also uses the MOT2017 data set and compares it with existing related tracking algorithms.It is concluded that the algorithm adopted in this chapter is in tracking accuracy and tracking accuracy.,Recognition accuracy and the total number of false alarms are more prominent,the tracking accuracy can reach 20.3%,and the tracking accuracy can reach 78%.Based on the analysis of the test set,it can be concluded that compared with the previous algorithms based on deep learning,the improved algorithm has outstanding performance in tracking accuracy and tracking accuracy.At the same time,the optimized algorithm also has better performance.Better robustness.