巡检机器人中的指针式仪表读数识别系统

为了解决智能巡检机器人仪表读数识别中易受光照变化影响、识别精度不高等问题,结合高压变电场中常见指针区域的图像特点,建立了指针式仪表读数高精度识别系统。鉴于巡检机器人的室外工作环境,提出了迭代最大类间方差算法,实现了多种光照条件下仪表图像的指针区域提取。经过分析指针转动和图像特性,提出了基于Hough变换的指针角度识别,推导出指针角度与仪表读数的函数关系。该算法较传统Hough变换角度提取法,增加了指针中心线通过表计中心等约束条件,提高了指针角度提取的精度,降低了搜索数据量和搜索时间。通过大量实验验证所建立的表计读数识别系统可实现室外各种光照条件下表计读数识别,获得95%以上的正确识别率。多组鲁棒性实验分析表明,该系统对光照条件、指针宽度、表盘干扰、拍摄角度(不产生指针阴影)具有较好的鲁棒性,但由于拍摄角度而产生指针阴影时,会引起较大的指针中心线提取和角度计算偏差,从而降低仪表读数识别精度。

Reading recognition system of pointer meter used in inspection robot

In order to solve the problems that the performance of instrument reading recognition system in intelligent patrol robot is easily affected by varying illumination and the recognition accuracy is low, a high accuracy reading recognition system for pointer instrument is built according to the characteristics of pointer area image commonly found in high voltage electronic field. In view of the outdoor work environment of patrol robot, an iterative maximum between class variance algorithm is proposed in this paper to achieve the extraction of pointer area in instrument image under various kinds of illuminations. After analyzing the pointer rotation mechanism and image characteristic, the pointer angle recognition algorithm based on Hough transform is proposed, the function relationship between pointer angle and instrument reading is deduced. Compared with traditional pointer angle extraction algorithm based on Hough transform, the constraint condition of pointer center line going through the meter center is introduced in the algorithm, which improves the accuracy of pointer angle extraction, and reduces the amount of search data and search time. Plenty of experiments verify that the instrument reading recognition system proposed in this paper can achieve meter reading recognition under various kinds of outdoor illumination conditions and achieve a correct recognition rate better than 95%. The analysis results of various robustness experiments show that the proposed system has good robustness to illumination condition, pointer width, dial interference, imaging shooting angle (without the shadow of pointer). However, when the shadow of pointer is generated due to imaging shooting angle, large bias will occur in the extraction of pointer central line and calculation of angle, which decreases the accuracy of the meter reading recognition.