The invention belongs to the technical field of elevator monitoring. The invention aims to provide the elevator trapped person detection method based on video analysis, and the elevator trapped person detection method is high in real-time performance and self-adaptive to multiple scenes so as to reduce the workload of manual monitoring. According to the technical scheme, the invention discloses an elevator trapped person detection method based on video analysis. Comprising the following steps: step 1, manually marking passenger, elevator door and floor readings of each frame of image in an elevator car monitoring video, and training by adopting a deep convolutional neural network by taking the passenger, elevator door and floor readings as a data set to obtain a multi-task detection model LiftDet for passenger target tracking, elevator door state judgment and floor readings identification; step 2, for a to-be-detected video, obtaining a detection set Ro by using a multi-task detection model LiftDet; and step 3, selecting a part in which people are not trapped in the elevator car monitoring video, and evaluating a standard limit frame number in which the passengers normally stay.
本发明属于电梯监控技术领域。目的是提供一种基于视频分析的电梯困人检测方法,实时性强、自适应多场景的电梯困人检测方法,以减少人工监控的工作量。技术方案是:一种基于视频分析的电梯困人检测方法;包括如下步骤:步骤1:对电梯轿厢监控视频中每一帧图像的乘客、电梯门与楼层示数进行手动标注,并以此为数据集采用深度卷积神经网络进行训练,得到乘客目标跟踪、电梯门状态判断以及楼层示数识别的多任务检测模型LiftDet;步骤2:针对待检测的视频,用多任务检测模型LiftDet得到检测集合Ro;步骤3:选取电梯轿厢监控视频中未发生困人异常的部分,对乘客正常停留的标准极限帧数进行评估。
Elevator trapped person detection method based on video analysis
一种基于视频分析的电梯困人检测方法
2023-08-04
Patent
Electronic Resource
Chinese
IPC: | B66B Aufzüge , ELEVATORS |
Elevator trapped person detection device and detection method thereof
European Patent Office | 2024
|Elevator trapped person detection system and control method thereof
European Patent Office | 2022
|