基于扩展扫描区域的现场表计指针识别方法

为了满足变电站巡检机器人对现场表计指针的智能识读要求,文中提出一种基于扩展扫描区域的现场表计指针识别方法。通过改进的SIFT算法进行图像匹配并求出单应矩阵,利用单应矩阵将在模板图像上标定的刻度弧线映射到待测图像上,通过扩展扫描区域,依据检测曲线路径上像素点的灰度值和确定指针位置并计算读数。文中对变电站现场的五种仪表,共计500张图片进行了测试。实验结果表明,文中所提算法简单实用、稳定且识别精度高,可以对多种不同的仪表进行识别,适用性强,在变电站多类指针表计识别中具有较好的应用前景。

Pointer recognition method based on extended scan area

To satisfy the intelligent identification requirements of the substation inspection robot for the field meter pointer, a pointer recognition method based on the extended scan area has been proposed in this paper. Image matching is performed by the improved SIFT algorithm and the homography matrix could be obtained. The scale arc calibrated on the template image is mapped onto the test image by the homography matrix. Expanding the scanning area, the sum of the gray values of the pixel on the path of the detection line is determined. It determines the position of the pointer and calculates the reading. Five meters have been tested in this paper, and a total of 500 pictures. The experimental results show that the proposed algorithm is simple, practical, stable and has high recognition accuracy. It can identify many different meters and has strong applicability. It has higher practicability in multi-class pointer meter identification.