Design and Development of an Identification System for Gauge Meter Value
Yen Jen Chen, Chun Yuan Huang · 2023
In many factories in Taiwan, there are still analog instruments used to monitor machine data. Traditionally, patrol personnel were required to walk around and manually record the data. However, this approach often led to discrepancies between the instrument data on the manufacturing site and the data returned by the instrument, or errors caused by personnel negligence in recording the data. In this study, we propose an automated recognition system for instruments. With this system, patrol personnel only need to use our inspection camera to capture images, and the data can be recognized in real-time and transmitted to the database. First, we use YOLOv5 object detection to outline the features of the instrument, then the second stage of YOLOv5 outlines whether the pointer is leaning to the right or left, and the last, Hough Transform is used to draw the angle of the pointer and convert it into instrument data using a formula. Our experimental data shows an accuracy rate of over 98%.