基于机器视觉的指针式仪表示数识别方法研究

Research on Recognition Method of Pointer Type Meter Based on Machine Vision

摘要:

基于机器视觉技术实现指针式仪表数据的自动读取具有重要意义，针对现有方法中存在的识别精度不高等不足，提出一种基于标定的指针式仪表数据视觉读取方法。首先，基于标定模板完成仪表表盘最大、最小刻度线识别与斜率计算；其次，通过仪表表盘图像预处理及连通区域筛选得到指针大致区域；然后，融合Hough和边缘聚类与拟合方法实现仪表指针边缘的精确定位，进而实现指针数据的识别与读取；最后，以某品牌避雷器监测器为例，对上述方法进行实验验证。结果表明，该方法能够准确、稳定的识别出指针式仪表读数。

Abstract:

Automatic recognition of the pointer type meter based on the machine vision is of great significance. Aiming at the shortage of low recognition accuracy in the existing methods, a visual reading method based on calibration for pointer type meter is proposed. Firstly, the largest and smallest scale lines of the instrument dial are recognized and their slopes are calculated according to the calibrating template; Secondly, the approximate region of the pointer is obtained by preprocessing the image of instrument dial and filtering the connected regions. Then, the Hough and edge clustering and fitting method are integrated to get the precise position of the edge of the instrument pointer, and then the recognition and reading of the pointer data are realized. Finally, an arrester monitor of a certain brand was used to verify the above method experimentally. The results showed that the pointer reading could be recognized accurately and steadily with this method.