Research on Identification and Positioning Method of Coal Carriage Based on Improved YOLOv5

Nowadays, some power plants have gradually begun to use mechanical automatic residual coal cleaning devices to clean up the remaining coal in the coal transportation carriage, and the key to realizing the automatic cleaning of the remaining coal is to accurately identify and locate the coal transportation carriage; In order to solve this problem, a method for identifying and positioning coal carriages based on improved YOLOv5 and monocular vision was proposed, and the feature extraction ability of the coal carriages was improved by integrating a convolutional block attention module (CBAM) into the model. Therefore, the backbone network in the YOLOv5 module was replaced with a lightweight MobileNetv3 network to reduce the computational cost of the model and improve the detection speed of the algorithm. By obtaining the pixels of the corner points in the candidate frame, the accurate positioning of the coal carriage was realized by using the spatial geometric relationship. Experimental results show that the mAP of the improved YOLOv5 algorithm reaches 98.3%, which is 4.6% higher than that of the original algorithm. After fixing the industrial camera, the maximum positioning error is between ±15mm, which can meet the operation of the mechanical automatic residual coal cleaning device into the carriage.