基于双目深度图像的自动扶梯乘客危险行为识别与预警系统

A Passenger Dangerous Behavior Recognition and Early Warning System of Escalator Based on RGB-D Sensor

基于双目深度视觉传感器,设计了一种融合双目立体匹配算法和深度神经网络模型的方法,针对乘客危险行为识别与预警的需求,构建了一套可应用于自动扶梯场景的乘客危险行为识别与预警系统。实验结果表明,该系统能有效识别乘客摔倒、逆行、探头、出入口滞留以及携带婴儿车和大件行李等多种危险行为,并能自动预警。

Based on the binocular depth(RGB-D)sensor,a method of integrating binocular stereo matching algorithm and depth neural network model is designed.Aiming at the needs of passenger dangerous behavior recognition and early warning,a set of passenger dangerous behavior recognition and early warning system which can be applied to escalator scene is constructed.The experimental results show that the system can effectively identify various dangerous behaviors,such as passenger falling,retrograde,probe,entrance and exit detention,carrying pram and heavy luggage,and can automatically warn.New ideas and solutions for improving the operation safety of escalators are provided.