Search

In view of the problems such as the large workload of video monitoring on drilling site and the lack of effective utilization of massive video data, firstly, the risk-based video monitoring layout optimization method was put forward by using particle swarm optimization algorithm after comprehensive consideration of the monitoring area coverage, key monitoring in high-risk areas and cost and other factors. Based on the identification and screening of high-risk behaviors with high accident consequences, the overflow monitoring scene was selected to design a deep neural network based on Faster RCNN and OpenPose frame to identify the arrival of personnel and squat sampling actions. Video intelligent analysis technology was developed, and video intelligent analysis and alarm system was developed to carry out real-time behavior detection of onsite overflow monitoring process. The results show that the human sampling motion recognition accuracy is 85.6%, and the system detection rate is 130 ms/frame, achieving good practical results.