Research on Underground Personnel Behavior Detection Algorithm Based on Lightweight OpenPose

To effectively identify the unsafe behaviour of underground workers in coal mines, this paper designs an intelligent detection system of underground workers’ behaviour based on a lightweight OpenPose algorithm. Firstly, the system obtains the coordinates of key points of human bones, detects the posture of falling, climbing and pushing by constructing different detection algorithms, and deploys the posture detection model to the industrial computer; Secondly, the industrial computer reads the video of the infrared camera and transmits the information of the detected unsafe behaviour to the alarm. The experimental results show that the detection speed can reach the actual application effect of single-channel 30fps or dual-channel 20fps.