Multi-sensor fusion-based intelligent supervision system for coal storage yard operation safety

Workers and vehicles are the key objects of coal yard safety management. It is of great significance to obtain high-precision positioning information in real time to avoid safety accidents. Studying the detection and positioning method of personnel and vehicle based on ultra-WB and visual integration, according to the scene characteristics of the closed coal yard, the method based on flight time-time arrival difference is proposed to realize the high-precision positioning of label targets in the coal yard; In order to solve the interruption of UWB location caused by occlusion, UWB detection and identification method based on Transformer and Yolov7 is proposed to estimate the position in the image, and the Kalman filter method is used to realize the fusion positioning. This method was tested in the coal yard of Ningxia Lingwu Power Plant of National Energy. By collecting the data images of operation scenes in the coal field, the data set of operators and vehicles was made. The average accuracy index of the target detection model [email protected] is 0.925, and the image detection speed of the model deployed on the site is 30 frames per second. The experimental results of the detection model show that the model has high detection accuracy and fast detection speed for the main motion targets in the coal storage field, and meets the actual requirements of the coal field application in real time. The comparative experimental results show that the proposed target detection model improves the [email protected] by 1.4% over the original Yolov7 without sacrificing the processing frame rate. The positioning accuracy range in the coal yard is 0.5～0.9 m, and the positioning refresh frequency is 10 frames per second, which effectively solves the problem of UWB positioning interruption caused by occlusion.