A Comparative Study of Meter Detection Methods for Automated Infrastructure Inspection

TL;DR

This paper compares shape-based, texture-based, and background information-based methods for meter detection in images captured by autonomous inspection robots, highlighting the superior performance of background information-based techniques in various conditions.

Contribution

It introduces and evaluates three distinct meter detection methods, demonstrating the effectiveness of background information-based detection for diverse meter shapes and sizes.

Findings

Background information-based method detects farthest meters regardless of shape.

It reliably detects meters with a diameter of 40 pixels.

The method outperforms shape and texture-based approaches in varied scenarios.

Abstract

In order to read meter values from a camera on an autonomous inspection robot with positional errors, it is necessary to detect meter regions from the image. In this study, we developed shape-based, texture-based, and background information-based methods as meter area detection techniques and compared their effectiveness for meters of different shapes and sizes. As a result, we confirmed that the background information-based method can detect the farthest meters regardless of the shape and number of meters, and can stably detect meters with a diameter of 40px.