Abstract: For machine vision based calibration systems of pressure gauges with pointers, a scheme based on the frame difference method and the angle method is presented for automatic reading recognition of index instruments. A method that uses the extreme points in detected regions to determine the centroids of pointers is also proposed to solve the problem of centroid deviation for incompletely detected pointers. First, the center of the dial plate is determined and the region of the dial plate is segmented using Hough circle detection. Then the pointer region is detected using the three-frame-difference method. The pointer centroid is located by selecting the two points closest to the diagonal vertexes of the minimum rectangle enclosing the pointer region from the extreme points in eight directions. Finally, the deflection angle of the pointer is computed and the pointer reading is recognized via connecting the pointer centroid and the center of the dial plate. Experimental results reveal that the regional extreme points based pointer centroid location can simply and effectively correct the centroid deviation for incompletely detected pointers. The overall scheme can accurately recognize the readings of pressure gauges with pointers.

Key words: pointer instrument, reading recognition, frame difference, centroid extraction

摘要： 针对基于机器视觉的指针式压力表校验系统，给出了一种采用帧差分法和角度法自动识别指针仪表读数的方案，并且为了解决残缺指针区域的质心偏移问题提出了一种利用区域极值点确定指针质心的方法。首先基于霍夫圆检测进行表盘中心定位和表盘分割；然后采用三帧差分法检测指针区域，并从其八方向极值点中选择最接近指针区域最小外接矩形对角顶点的两点来定位指针质心；最后连接指针质心和表盘中心以计算指针偏转角度和识别读数。实验结果表明基于区域极值点的指针质心定位可以简单有效地修正残缺指针区域质心偏移，整体方案能够较准确地识别指针式压力表读数。

关键词: 指针式仪表, 读数识别, 帧差分, 质心提取