Computer Vision System for Reading Analog Gauges at Power Substation

This paper proposes a computer vision system for reading analog gauges at a power substation. The gauges mentioned in this paper include oil level gauges, winding temperature gauges and SF6 gas density gauges. We present two different algorithms since an oil level gauge has different characteristics from a winding temperature gauge or an SF6 gas density gauge. For the oil level gauge, a color segmentation method is used to detect the pointer. Then we use Canny Edge Detector in combination with the Hough Circle Transform method to detect scale marks. Based on the position of the pointer and the marks, we detect the indicating value of the gauge. For winding temperature gauges and SF6 gas density gauges, we find the direction of the pointer based on the geometry properties of a double-edged pointer. With this information, we can calculate the value indicated on the gauge face using interpolation between the direction of the pointer and the value of the gauge. The system error is almost zero in case of an oil level gauge and remains low in the case of a winding temperature gauge and an SF6 gas density gauge. Mobile robots integrated with this computer vision system will work faster and more accurately than humans and will also help avoid dangerous accidents.