Image Processing for Automatic Reading of Electro-Mechanical Utility Meters

Abstract

Electro-mechanical meters are commonly employed to measure the consumption of utilities. Basically there exist two types of analog meters: the ones that use rotary dials (like an odometer) and the ones with pointer dials (like a speedometer). Former approaches to automated meter reading have dealt with the first kind of meters. Considering that automated reading of the latter ones can be confusing, in this work we introduce a methodology based on image processing and segmentation to enable the image acquisition and processing of pointer dials to obtain efficiently and accurately readings. This methodology uses an image acquired with a smart phone and by applying a sequence of image processing functions it finds and extracts the dial images of such meter images. Then the methodology identifies the position of the pointers followed by a clever implementation that enables the reading. The database is composed with more than a hundred images taken under different light conditions, perspectives and angles. The method is able to extract the reading in an average of 3 seconds, with a 92 % accuracy with images taken in-field. Our method, enables the use of a common smart phone to acquire and automatically extract the reading of a pointer-type dial meter. This allows interesting applications that could help people to monitor their energy consumption and learn patterns to save energy. This could be one step ahead of energy saving policies that can be discovered through massive data analysis.