基于视频的机场出发层违规接客车辆识别方法

Algorithm for identifying violation behavior of vehicles of picking up passengers at airport departure floor based on surveillance video

为有效监管机场出发层车道边车辆违规接客行为,降低违规接客行为对陆侧交通通行能力的影响,建立基于监控视频自动识别航站楼出发层违规接客车辆的方法,对违规接客车辆的行为特征进行分析,提出基于YOLO_v4&深度简单及时跟踪(DeepSORT)的车辆运动状态检测算法与车辆接客行为识别算法。首先,使用YOLO_v4识别目标,获取目标的类别与位置信息,统计车辆目标在各个运行状态下的位移数据,分析车辆目标运行状态阈值,建立基于固定监控机位的车辆运动状态检测算法。然后,结合识别跟踪的信息,分析发生接送客行为时各目标间的行为关系与发生区域,以车内人数变化为区分接客与送客行为的重要依据,建立了基于YOLO_v4&DeepSORT识别跟踪结果的接送客行为检测算法。其中,违规识别算法使用YOLO_v4和DeepSORT识别、跟踪、记录并处理目标类别与位置信息,判断车辆的运行状态与驾乘人员在车辆附近的相关行为;在车辆停止时记录乘客行为信息,计算车内人数变化情况,在车辆消失于监控区域时根据最终车内人数变化情况判断车辆的接送客行为。最后,使用Python语言实现机场出发层违规接客车辆识别算法,并以昆明长水国际机场2019年9月1日~3日的监控视频进行违规识别算法测试。结果表明:提出的违规接客识别算法从机场出发层识别接客行为较为有效,其识别准确度达到了86.49%,可为机场出发层违规接客识别提供有效监控手段;同时,其中仅有0.41%的车辆被误判为违规接客,具有较低的误识别率,该算法可较好地区分接客行为与送客行为,有助于在接送客混行的出发层中将正常车辆与违规车辆加以区分。

To effectively monitor the violation behavior of vehicles that pick up passengers at the departure floor of airports,and reduce the impact of such violation behavior on the curbside traffic capacity,a method based on YOLO_v4&DeepSORT algorithm was proposed,to automatically identify vehicles that pick up passengers at the departure floor based on surveillance video.Firstly,YOLO_v4 was used to detect the targets,the displacement of vehicle targets in each running state was recorded,and the threshold value of the vehicle targets’running state was analyzed.Based on the fixed monitoring position,a vehicle motion state detection algorithm was developed.Then,based on the identification and tracking information obtained by YOLO_v4&DeepSORT,an identification algorithm for pick-up and drop-off behaviors was established.The algorithm used YOLO_v4&DeepSORT to identify the target and to record and process the classification and location information.After the running state of the vehicle and the related behaviors of the driver and passenger near the vehicle were identified,the passenger behavior information could be recorded when the vehicle stops,and the change in the number of people in the vehicle could be calculated.After the vehicle disappears from the monitoring area,the pick-up behavior of the vehicle could be identified based on the final change in the number of passengers in the vehicle.Finally,the algorithm was programmed by Python and tested using the surveillance video of Kunming Changshui International Airport.The results show that the proposed algorithm can effectively identify the pick-up behavior from the departure floor of the airport,and the identification accuracy of the algorithm reaches 86.49%,which can provide an effective way to identify the violation pick-up behavior at the departure floor.At the same time,only 0.41%of vehicles are misidentified,which shows a low misidentification rate,indicating that the algorithm is helpful to differentiate normal vehicles and violation vehicles among all the vehicles on the departure floor.6 tabs,12 figs,24 refs.