Automatic Electricity Meter Reading Based on Image Processing

Abstract

Electricity meter reading is an important but time-consuming task in utility management, often relying on manual data collection. This manual method not only increases the risk of human errors but also leads to inefficiencies, potential data tampering, and difficulty accessing meters in densely populated or hard-to-reach areas. To address these challenges, we propose an automated solution for electricity meter reading using advanced image processing techniques combined with machine learning models. The aim of this research is to streamline the meter reading process, improve accuracy, and ensure fair billing for customers. The proposed system uses state-of-the art algorithms, including YOLO V8 (You Only Look One, Version 8) for real-time object detection and fast OCR (Optical Character Recognition) for digit recognition, to extract meter readings from images. We also implement techniques specifically designed to handle small objects in images, allowing the system to accurately read meter digits, even under different environmental conditions such as changing lighting and different meter types. The dataset consists of multiple images of electricity meters captured in different real-world conditions. By processing these images, our system can automatically detect and interpret meter readings, overcoming challenges such as poor lighting and different meter designs. The system’s performance was evaluated, achieving 91% accuracy in reading and interpreting electricity meter digits. This solution not only improves operational efficiency for utility providers but also ensures transparency and fairness in the billing process for consumers. Our results demonstrate that this approach can significantly increase the accuracy and speed of meter reading operations, reduce errors associated with manual reading, and contribute to the modernization of utility management systems.