Development of a Public Transport Passenger Flow Statistics System Based on Artificial Intelligence
Aiming at the current problems in bus passenger flow statistics,such as accuracy and efficiency,target occlusion,false detection,and missed detection,a bus passenger flow statistics system based on YOLOv5 object detection algorithm combined with DeepSORT multi-objective tracking algorithm is pro-posed based on real-time monitoring video data inside the bus.Firstly,a method based on detecting and tracking human head features was selected to statistically analyze and screen video data from different time periods,weather conditions,and lighting intensities to ensure data diversity and representativeness.The human head was annotated using the AutoLabelImg semi-automatic annotation tool.The YOLOv5 algo-rithm can obtain the detection box and confidence level of the target object,while the DeepSORT algorithm can obtain the ID of the target object,solving the problem of identifying the weight of passengers.Finally,the counting of the target object is completed through the point crossing line counting method.After exper-imental testing,the proposed bus passenger flow statistics system achieved a detection accuracy of 95％,a recall rate of 94％,and mAP:0.5 value of 0.97,greatly improving the accuracy of passenger flow statis-tics.
bus passenger flow statisticstarget detectiontarget trackingYOLOv5DeepSORT
NETL
NSTL
万方数据
