摘要
为了满足晶硅光伏电池视觉印刷高精度边缘定位要求,本文提出了一种改进的亚像素边缘定位算法。该算法首先应用Sobel算子和线性插值得到边缘垂直方向上新的边缘图像;最后对插值的边缘图像使用改进的灰度矩算子,得到的亚像素边缘。同时本文进行改进亚像素检测算法与传统亚像素检测算法的精度及鲁棒性对比实验。结果表明:本文算法比传统亚像素定位算法有着更好的定位精度和鲁棒性,直线检测精度能达到1μm;在高噪声污染的图像有着较高的检测精度,其标准差为0.195μm。
To realize high accuracy and high stability of edge detection on the automated vision screen printer of crystalline silicon solar cells,a new edge detection algorithm based on gray moment proposed by Tabatabai is proposed.Frist,SOBEL operator and line fitting are used to find edge in pixel accuracy so that a new image is obtained by bilinear interpolation along the vertical direction of the fit line.Finally,the improved gray-moment operator is used to get the sub-pixel edge in the new image.At the same time,comparative experiment of accuracy and stability is implemented between improved operator and traditional operator.Experiment results indicate that operator proposed here detects edge more accurate and stable.Edge detection accuracy is less than 1,and standard deviation of edge detection is less than 0.195μm in high noise pollution image.
出处
《计量与测试技术》
2021年第7期44-47,共4页
Metrology & Measurement Technique
关键词
灰度矩
亚像素
边缘检测
直线拟合
gray moments
sub-pixel accuracy
edge detection
line fitting
