基于图像取反的同心圆环半径和圆心提取方法  

The method to extract concentric ring radius and circle center based on image inversion

作　　者：欧阳烨锋 崔建军[2] 张宝武 陈恺 杨宁 方振远 OUYANG Yefeng;CUI Jianjun;ZHANG Baowu;CHEN Kai;YANG Ning;FANG Zhenyuan(College of Metrology and Measurement Engineering,China Jiliang University,Hangzhou 310018,China;National Institute of Metrology,Beijing 100013,China;Shaanxi Institute of Metrology Science,Xi’an 710100,China)

机构地区：[1]中国计量大学计量测试工程学院,杭州310018 [2]中国计量科学研究院,北京100013 [3]陕西省计量科学研究院,西安710100

出　　处：《激光技术》2024年第1期135-139,共5页Laser Technology

基　　金：国家市场监督管理总局科技计划资助项目(2022MK220;2021MK188);浙江省科技厅基础公益研究计划资助项目(LY22E05005)。

摘　　要：为了解决非定域干涉同心圆环图像的圆心和半径的精确提取问题,提出了一种基于图像取反的暗亮环兼顾的算法,以点光源非定域干涉同心圆环为对象,在平滑降噪预处理基础上,利用霍夫圆变换获得参考圆心,以参考圆心为基础分别在其x方向和y方向上[-50,50]像素范围内每隔5 pixel取一条强度轮廓线,每一条强度轮廓线取反后,分别实现暗环和亮环峰位坐标的精确识别,再利用圆回归拟合得到系列圆心和半径,通过系列圆心和半径的平均运算实现图像最终圆心和半径的精确提取。结果表明,靠近中心的3个亮环和3个暗环各自的圆心坐标非常接近,相对于总平均值的最大偏差为-3.7 pixel,即相对于总平均值的最大相对偏差为-0.15%,重复性很好;相邻两个圆环半径平方比值相对于实际比值的误差在-4.18%~0.36%之间变化,验证了算法的可行性。该研究对实现自动化检测、提高测量精度是有帮助的。Aiming at the problem of accurate extraction of circle center and ring radius in non-localization interference concentric ring image,taking point light source non-localization interference concentric rings as the object,an algorithm based on image inversion for simultaneous caring of both dark and bright rings was proposed.The reference center was obtained using the Hough circle transformation after the pre-processing of smooth noise reduction.Then intensity outlines with an interval of 5 pixels in the range of[-50,50]pixels in its x-direction and y-direction,respectively,were taken.With the inversion of each intensity contour line,accurate identification of dark and bright ring peak coordinates,respectively,was achieved.Finally,the accurate extraction of the circle center and radius of the image was achieved by averaging multiple circle parameters after circle regression fitting of obtained series circle center and radius.The results show that the center coordinates of the three bright rings and the three dark rings are very close to each other and have good reproducibility with the maximum deviation of-3.7 pixels to the mean value,i.e.,a maximum relative deviation from the mean is-0.15%.The relative error of the radius-square ratio of two adjacent rings to the actual ratio varies between-4.18%and 0.36%,verifying the feasibility of the algorithm.This research is helpful in realizing automatic detection and improving measurement accuracy.

关 键 词：图像处理 非定域干涉 霍夫圆变换 图像取反 同心圆环 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术] TH691.9[自动化与计算机技术—计算机科学与技术]