基于图像识别的多指针仪表自动读数方法

Image recognition based automatic reading method for multi-pointer meter

针对多指针仪表的指针读数识别问题,提出了一种基于色差变换的定位方法和基于表盘结构特征的指针读数自动识别与判读方法.采用LRCD变换、数学形态学处理方法和p-分位法完成有效区域筛选及水表的指针定位,根据表盘中心到指针回转中心的连线与子表盘0刻度线的夹角等特征,计算出子表盘0刻度线与指针指向线段的夹角,进一步识别并判读出指针读数.实验结果表明,该定位及识别算法计算简单,具有较高的准确率,克服了表盘随机角度倾斜对读数识别算法的影响.

Aiming at the recognition of pointer reading for multi-pointer meter,a positioning method based on chromatism transformation as well as an automatic recognizing and reading method based on structural characteristics of a dial were proposed.The luminance and red color difference（LRCD） transform,mathematical morphology based processing method and p-tile method were used to accomplish the selection of effective area and the pointer positioning of water meter.According to such features as the angle between the linking line of dial center to pointer rotatory center and the zero graduation mark of sub-dial,the angle between the zero graduation mark of sub-dial and the pointer directed line was calculated,and furthermore,the pointer reading was recognized and read.The experimental results show that the proposed algorithm for positioning and recognizing has calculating simpleness and higher accuracy,and can avoid the effect of random angle tilt of dial on reading recognition algorithms.