A Pointer Meter Recognition Algorithm Based on Deep Learning

Abstract

In order to build an intelligent, unmanned and managed substation, the substation gradually adopts inspection robots instead of manual work. However, during the inspection process, the automatic identification of the pointer type meter always has the problem that the recognition accuracy is not high and is susceptible to illumination changes. In this paper, the problem of pointer instrument identification is studied. The method of instrument detection and localization algorithm, pointer and instrument scale fitting in complex environment is studied systematically. A pointer meter identification algorithm suitable for intelligent substation inspection robot is proposed. Using the positioning function of deep learning, the instrument classification algorithm based on Faster R-CNN is used to classify three types of tables: voltmeter, ammeter and digital table. Image processing of the positioned pointer meter image, such as tilt correction, extraction of scale segments, pointer line segments, line fitting, repairing the default tick marks, etc. is operated to calculate a more accurate pointer reading.