气门几何尺寸的多种边缘高精度检测  

High-precision Multi-edge Measurement Method for Geometrical Size of Valve

作　　者：黄源[1] 徐杜[1] 蒋永平[1] 

机构地区：[1]广东工业大学信息工程学院,广州510006

出　　处：《科学技术与工程》2015年第2期93-96,103,共5页Science Technology and Engineering

基　　金：广东省教育部产学研结合项目(2010B090400382)资助

摘　　要：目前边缘检测算法只能检测水平边缘、垂直边缘,且检测精度低、处理速度慢、抗噪性能差;针对上述存在的缺陷,提出一种气门几何尺寸的多种边缘高精度尺寸检测算法。首先采用中值滤波和高斯滤波对气门采集图像进行预处理,然后针对不同的边缘使用不同检测算法实现图像边缘的像素级定位。在像素级边缘定位的基础上采用几何质心法亚像素边缘定位实现图像边缘的亚像素级精确定位。最后采用畸变校正技术对图像中边缘像素点的坐标进行校正,得到没有畸变情况下边缘像素点的理想坐标,根据像素当量计算得到气门的各个尺寸。通过在光学图像检测系统中的实际应用,证明提出的算法精确且稳定,满足高精度视觉检测的要求。Currently only the edge detection algorithm to detect the horizontal edges, vertical edges, and the de- tection accuracy is low, slow processing speed, poor anti-noise performance for the above defects, a variety of valve geometry edge precision size detection algorithm was presented. Firstly, the median filtering and Gaussian filter to capture images of the valve pretreatment, and then a different edge detection algorithms for different positioning to achieve pixel-level image edge was used, using sub-pixel edge geometric centroid based on the pixel level edge po- sitioning positioned to achieve sub-pixel image edge accurate positioning, and finally distortion correction technology edge pixel coordinates of the image correction to obtain the coordinates of the ideal case of no edge pixel distortion was used, based on the size of no edge pixel distortion, based on the size of each pixel to obtain equivalent valve. Through practical application in optical image detection system, the proposed algorithm proves accurate and stable, to meet the requirements of high-precision visual inspection.

关 键 词：边缘检测 滤波 亚像素 畸变校正 

分 类 号：TN911.73[电子电信—通信与信息系统]