The indication reading recognition method of pointer-type pressure gauge based on image processing

Abstract Reading recognition for pointer-type pressure gauges based on computer vision (CV) is significant for realizing unmanned monitoring and automatic calibration. Typical recognition methods significantly rely on identifying the scale ring and the pointer because they primarily obtain the reading from the angular displacement of the pointer. For widely used economical pressure gauges, significant recognition errors can be caused by certain inherent defects, such as the misalignment of the pointer’s rotation center (RC) and scale center, which are insensitive to human eyes but fatal to CV. Additionally, dial stains may lead to the misidentification of scale marks. This paper proposed a novel reading recognition method for pointer-type pressure gauges that founded the scale system on a unified point for the scale center and the pointer RC. The scale center is identified utilizing the geometrical morphology and distribution characteristics of certain scale marks, and then it is taken as the basis to recognize the remaining scale marks and build the scale system. With the pointer identified from the interfering background, its RC is fitted to the scale center by generating its indicating vector based on the position of the pointer tip and the identified scale center. The implementation results proved the validity of the recognition method, and the reading accuracy met the requirement for an automatic calibration system.