MAMRS: Mining Automatic Meter Reading System Based on Improved Deep Learning Algorithm Using Quadruped Robots

Abstract

The meter reading work in the power distribution room of a mine is traditionally carried out by manual inspection, which requires several workers and often lacks timeliness and efficiency. Owing to the widespread use of quadruped robots, this paper proposes a Mining Automatic Meter Reading System, named MAMRS, to identify and read pointer and digital meters in mine distribution rooms using quadruped robots. The new method consists of three stages. Initially, this technique identifies the type of meter using the ResNet18 convolutional neural network model, then proceeds to extract the dial area image using the YOLOv5 object detection algorithm, and finally reads the pointer and digital meters’ data using the U2-net algorithm and SVM, respectively. Experimental results show that the recognition accuracy rates for meter classification, pointer meter reading, and digital meter reading are 99.87%, 85.35%, and 90.73%, respectively. The MAMRS fulfills the need for fully automated and intelligent inspection of mine distribution rooms, resulting in significantly enhanced accuracy, flexibility, and innovation in inspection processes. Moreover, it reduces labor costs and improves inspection efficiency. The findings of this study serve as a reliable reference for intelligent inspection practices in the power distribution rooms of metal mines.

Keywords

• deep learning
• mine power distribution room
• quadruped robot
• meter reading
• fully automatic inspection