Research on Behavior Identification Method for Construction Personnel in Live Areas of Power Grid Based on Target Tracking
A target tracking based personnel behavior identification system was designed to address the difficulties in monitoring the behavior of construction personnel in live areas of the power grid.Combining computer vision technology and sensor technology,bi-directional feature pyramid network(BiFPN)is used to optimate the Loss function to improve YOLOv5s,and introduces Kalman fil-ter model to realize target tracking of construction personnel's behavior.The results show that the identification accuracy in the NEU Pole and Baidu FSG datasets is over 92％.When the accuracy is 0.9,the corresponding recall rates are 0.82 and 0.84,respective-ly.The research system correctly identified 13 and 44 annotations in two scenarios,respectively.By comparing the above data,it can be seen that the research system can accurately identify and track the behavior of construction personnel in live areas of the power grid in real-time.Keywords:YOLOv5s algorithm;Power grid construction;Identification of personnel behavior;Target tracking algo-rithm;Kalman filter algorithm;Decision Tree Algorithm.
YOLOv5s algorithmidentification of personnel behaviortarget tracking algorithmkalman filter algorithm
万方数据
维普
