A Novel Method for Recognizing Readings in Images of Electric Circuit Inspection Meters Based on Deep Learning

Abstract

For electric circuit inspection, the conventional manual inspection method has a series of problems including the heavy work load of recording the readings, the low accuracy, and the hidden safety hazards. The intelligent recognization method for images of meter readings based on digital image technology has a great practical value. However, the existing recognization methods for meter readings based on deep learning generally ignore the extraction of key points such as the pointer and scale on meter dashboard, the existing algorithms are of poor robustness and anti-jamming ability, therefore, this paper aims to study a novel method for recognizing readings in the images of electric circuit inspection meters based on deep learning. At first, this paper corrects the tilt of meter dashboard, and accurately positions the dashboard center. Then, based on the YOLOv5 network model, this paper constructs the said recognization model, gives the structure of the YOLOv5 network model, and introduces its working principle. At last, experimental results are drawn to verify the validity of the proposed method for processing the images of meter readings and the constructed recognization model.