Research on the method of identifying the top plate pallet in the underground roadway of coal mine

Pallet is an important feature of mine intelligent application based on machine vision technology such as roadway roof monitoring and equipment automatic positioning, and the accurate identification of pallet is the key to this kind of application technology. A roof pallet identification method based on improved Faster R-CNN is designed. Firstly, ResNet50 network with better feature extraction effect is selected as the backbone network. Secondly, the K-Means ++ clustering algorithm is used to optimize the candidate frame size in the regional suggestion network, so that the model can accurately locate the frame. Finally, the attention mechanism module is introduced in ResNet50 to enhance the salience of the pallet in the image, enhance the network's attention to the pallet, and then improve the detection efficiency and accuracy of the pallet. In the simulated roadway, the AGV car is used to simulate the boring machine, and the images of the pallet are collected by the airborne camera. The experimental results show that the improved Faster R-CNN model has good robustness, and the average accuracy (mAP) of the identification pallet is 92.53%, which is 7.59% higher than the original Faster R-CNN network.