高精度直线检测算法研究与误差分析  

Research on high-precision line detection method and error analysis

作　　者：袁继栋[1] 阳波[1,2] 郑煜[1] 段吉安[1] 

机构地区：[1]中南大学现代复杂装备设计与极端制造教育部重点实验室,长沙410083 [2]湖南师范大学图像识别与计算机视觉研究所,长沙410081

出　　处：《现代制造工程》2011年第2期19-23,共5页Modern Manufacturing Engineering

基　　金：国家自然科学基金重点项目(50735007);国家863高技术研究发展技术项目(2007AA04Z344)

摘　　要：针对图像处理中直线检测精度不足、运算速度慢等问题,对直线检测算法和精度进行研究,提出一种新的高精度直线检测算法。首先,基于任意两点计算直线参数值并以直线拟合的像素数目作为参数点的累加值;其次把累加值大于阈值的参数点保存在数据链表中并根据局部极大值或阈值判断直线;最后对算法的直线检测精度进行分析,建立直线拟合的误差模型。通过对0°到45°斜率角的直线检测进行误差分析发现,当直线段的长度为400像素时,斜率角的典型误差小于0.03°,当直线段的长度为600像素时,斜率角的典型误差小于0.01°。新算法与Hough、最小二乘算法比较,不但精度高,运算速度快,而且受噪声影响小。In view of the fact that line inspection in image process is not accuracy enough and runs slowly,a new high-precision line detection algorithm is proposed on the research of the normal line detection algorithms and their accuracy theory.First of all,the line parameter value is calculated based on any two points of a image and the number of the pixels in the line which is decided by one parameter is accumulated as a accumulated parameter;the second,a data list is established that stores those parameters greater than a certain threshold,then those potential lines can be recognized for they are the local maximum values or greater than a threshold;finally,the error model of the new algorithm is formed after precision theory analysis.The simulation experiments in which lines tilting from 0° to 45° are detected by the new algorithm state that the typical angular error is less than 0.01°when the length of the line is limited to 600 pixels instead of that the error is less than 0.03°when the length is 400pixels,and the comparative experiment results reveal that the new algorithm have more outstanding quality,not only higher accuracy,faster computing speed,but also smaller influence by noise,than the Hough and the Least Square Method（LSM）.

关 键 词：直线检测 误差分析 视觉测量 最小二乘 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]