为解决工业生产中指针式油位计仪表示数识别易受拍摄角度和光照条件影响的问题，结合工业生产场景下指针区域的图像特点，提出一种基于刻度准确定位的仪表示数识别方法，仅使用刻度与指针这两个必要特征来实现表盘读数。该方法先对巡检机器人拍摄的仪表表盘图像进行预处理，然后粗识别出表盘中可能是刻度的区域，再通过细识别精确定位出刻度。针对传统仪表示数识别方法难以解决的拍摄角度较大时刻度拟合的圆心和指针中心线的关系并非一定成立的问题，文中采取了一种改进的圆心选取方法，更加适应不同拍摄角度下的计算，提高了读数的精度。实验结果表明，文中方法获得的仪表读数精度平均达到 99. 08%，算法可靠稳定。

In order to solve the problem that the reading recognition of pointer-type oil-level meter is easily affected by the change of shooting angle and illumination conditions，a new method for recognizing the readings based on ac-curate scale location was proposed by considering the characteristics of pointer in the industrial production． It only uses the scale and pointer to calculate the reading of meter． In this method，firstly the images of oil-level meter taken by the patrol robot was preprocessed; then all possible scale areas in the picture were recognized coarsely;and finally the scale was accurately located by fine recognition． In view of the problem that the traditional instru-ment recognition method is hard to solve the relationship between the center of the circle and the center line of the pointer is not certain to be established when the shooting angle is large，this paper adopted an improved method to select the center of the circle． This method is more suitable for the calculation under the influence of the shooting angle change，and can improve the reading accuracy． The experimental results show that the reading accuracy of the instrument obtained by this method is 99. 08% on average，and the algorithm is reliable and stable．