作者：戴宗贤¹, 莫洪波¹, 周勇¹, 张震¹, 尹爱军², 梁子晓²

作者单位：1. 重庆市计量质量检测研究院, 重庆 401121;
2. 重庆大学机械工程学院, 重庆 400044

关键词：视觉测量;尺寸测量;边缘检测;亚像素插值算法

摘要：

针对现有视觉测量中的检测代价高，精度低和速度慢问题，该文提出一种基于计算机机器视觉的紧密内插值亚像素测量方法。该方法基于线性插值算法的原理，结合常规边缘检测方法和图像的灰度曲线图，利用阈值分割和标准长度进行亚像素自适应阈值选择。为验证该方法的有效性，对标准量块长度进行测量实验，并分析测量系统的误差影响因素。相比于传统的Canny算子方法检测结果，该方法的平均测量准确度提升46.2%。实验结果表明该算法的测量精度较高，可以快速、精确地测量出物体的几何尺寸。

A sub-pixel interpolation vision measurement method based on graph gray level
DAI Zongxian¹, MO Hongbo¹, ZHOU Yong¹, ZHANG Zhen¹, YIN Aijun², LIANG Zixiao²
1. Chongqing Academy of Metrology and Quality Inspection, Chongqing 401121, China;
2. College of Mechanical Engineering, Chongqing University, Chongqing 400044, China
Abstract:  Aiming at the high cost, low accuracy and slow speed problems of existing vision measurements, this paper proposes a novel sub-pixel interpolation algorithm based on computer vision for geometry size measurement. At first this paper discusses the mathematical principle of linear interpolation algorithm. On this basis, combining with conventional edge detection methods and the image grayscale curve graph, a sub-pixel interpolation algorithm based on graph gray level is proposed, which is utilizing the threshold segmentation and standard length for sub-pixel automatic threshold selection. This method has the measurement characteristic of high precision and low time cost. Finally, for verifying the validity the method proposed in this paper is applied to measure the size of standard gauge blocks and the measurement error is analyzed. Compared to the traditional Canny operator method, the measurement accuracy of proposed method is higher than that of canny operator method with 46.2% improvement on average. The experiment results show that the proposed method is available and has high precision for geometry size measurement.
Keywords: vision measurement;geometry size measurement;edge detection;sub-pixel linear interpolation algorithm
2019, 45(5):33-37  收稿日期: 2018-03-30;收到修改稿日期: 2018-04-24
基金项目: 国防基础科研计划（JCKY2016209B008）
作者简介: 戴宗贤(1986-),男,安徽六安市人,工程师,硕士,研究方向为测量与测试技术,图像处理
参考文献
[1] 王鹏. 一种局部二值化方法及其应用[D]. 长春:吉林大学, 2017.
[2] 徐晶晶. 低质量文档图像的二值化算法研究[D]. 武汉:湖北工业大学, 2017.
[3] 张梅, 王涛. 一种工业恶劣环境下的图像二值化方法及FPGA实现[J]. 微型电脑应用, 2017, 33(7):69-72
[4] FISCHER J, PRIBULA O. Precise subpixel position measurement with linear interpolation of CMOS sensor image data[C]//The 6 IEEE International Conference on Intelligent Data Acquisition and Advanced Computing system:Data, Technology and Applications, 2011.
[5] ERIC P. Visual hyper acuity representation and computation of high precisio position information[J]. Computer Vision, Graphies and Image Processing, 1986, 33:99-115
[6] 杨剑, 大尺寸视觉测量精度的理论和实验研究[D]. 北京:北京邮电大学, 2010
[7] 申晓彦. 轴对称构件几何尺寸的图形测量[D]. 太原:中北大学, 2010
[8] 董鸿燕. 边缘检测的若干技术研究[D]. 长沙:国防科学技术大学, 2008
[9] 季虎, 孙即祥, 邵晓芳, 等. 图像边缘提取方法及展望[J]. 计算机工程与应用, 2004, 14:70-73
[10] 李了了. 工业现场字符识别方法的研究[D]. 合肥:合肥工业大学, 2003.
[11] 周衍超, 罗浩彬. 机器视觉技术在工业机器人的应用研究[J]. 装备制造技术, 2015(6):193-195
[12] 刘力双, 孙双花, 吕奶光. 视觉系统中调焦函数的选择方法研究[J]. 北京机械工业学院学报, 2008, 23(1):1-4
[13] 张广军. 机器视觉[M]. 北京:科学出版社, 2004:33-42
[14] HUECKEL M H. An operator which locate edges in digitized pictures[J]. Journal of the Association for Computing Machinery, 1971, 18(1):113-125
[15] 李艳, 彭嘉雄. 基于D2样条插值和LOG算子的亚像素边缘检测[J]. 华中理工大学学报, 2000, 28(3):77-79
[16] 邹宁, 柳剑, 周曼丽. 基于Zernike矩的边缘检测方法[J]. 华中理工大学学报, 1999, 27(5):1-2
[17] 梁燕, 刘文耀. 一种新型非线性边缘检测方法[J]. 光电子.激光, 2003, 14(12):1332-1335
[18] ZHANG H T, DUAN F J, YE S H. Study on a fast method for sub-pixel edge detection[J]. Acta Metrologica Sinica, 2002, 23(4):263-265, 270
[19] 赵敏, 邱宗明, 黄秋红, 等. 提高图像测量分辨率的一种方法[J]. 西安理工大学学报, 2002, 18(1):80-83
[20] 娄训志. 基于小波变换的图像亚像素检测应用研究[D]. 武汉:湖北工业大学, 2011:22-38