一种复杂指针式仪表的读数自动识别算法

摘要

For the pointer instrument with complex surface background and liquid filler,in order to get a more accurate reading,a novel and practical reading recognition method based on image processing technology is proposed. Based on this method in the integrated use of adaptive threshold iteration method for image binarization and edge extraction algorithm combined morphological method and the silhou-ette for image pre-processing,Hough transform circle detection and para space average center detection are adopted to determine the cen-ter of the instrument. And the method is adopted to find the pointer angle which is the mean method of bidirectional scanning based on traditional central projection,and the meter reading is calculated by using the angle method. The traditional central projection method and its improved one are compared and verified,and the error of the two methods are compared and analyzed. The experimental results show that the improved algorithm has stable performance,especially for the rotation of the pointer instrument recognition with high accuracy and strong real-time performance.%针对仪表表盘背景复杂且有液体填充物的指针式仪表,为了得到更加精确的读数,提出了一种新颖实用的基于图像处理技术的指针读数识别方法.该方法在综合运用自适应阈值迭代法对图像二值化及形态学与剪影法结合的边缘提取算法进行图像预处理的基础上,采用Hough变换圆检测和para空间均值圆心的方法确定仪表的圆心,创造性地采用了基于传统中心投影法的双向扫描求取均值确定角度的算法,并应用角度法计算出仪表指针读数.采用传统中心投影法及其改进算法进行了同步对比验证实验,并对两种方法的误差进行了对比评估和分析.实验结果表明,改进算法性能稳定,特别对于旋转的仪表指针识别具有较高的精度和较强的实时性.