Real-time detection and location of reserved anchor hole in coal mine roadway support steel belt

Based on the current coal mine roadway using supporting steel belt steel belt auxiliary anchor support, the location of the supporting steel belt anchor holes is mainly done manually; if the location is not accurate, there will be large safety hazards and other problems. An intelligent real-time detection and location method of anchor holes in coal mine roadway support steel belt based on deep learning model and depth camera is proposed. First, to reduce the influence of water mist and dust on the camera and improve the image quality of the camera, the image is pre-processed using contrast limited adaptive histogram equalization. Second, the YOLOv5s model is improved by adding SPD-Conv and coordinate attention mechanisms to improve the detection capability of the model. Third, a real-time depth map restoration method that fully preserves object edge features is proposed to avoid errors caused by areas with depth values of 0 in the depth map when locating anchor holes in combination with a depth camera. Finally, the improved YOLOv5s model proposed in this paper combined with the repaired depth map was used to achieve the detection and location of anchor holes in a laboratory simulated tunnel with a location error of less than 5 mm and an average FPS of 28.9. In summary, the real-time target detection and location method based on a deep learning model and a depth camera is feasible in an unstructured environment in coal mines.