基于改进Canny算法的LED晶片边缘提取

作者：智少丹，李建勇*，王 恒 日期：2010-09-08/span> 浏览：3258 查看PDF文档

基于改进Canny算法的LED晶片边缘提取

智少丹，李建勇*，王 恒

（北京交通大学 机械与电子控制工程学院，北京 100044）

摘要：为了提高固晶机对LED晶片边缘提取的自动化程度和边缘质量，避免固定阈值Canny所必须的人工调整，增大检测方法的适应范围和稳定性，提出了一种基于大津法的Canny边缘检测算法进行晶片边缘提取。在分析了LED晶片的图像灰度分布特征之后，通过求得最大类间方差找到了自适应的分割阈值，并将其作为Canny算法的低阈值进行边缘的检测提取。实验结果证明，这种方法适合LED晶片的边缘提取，能够快速得到较为精确的晶片边缘轮廓和晶片两极边缘，对光线变化的鲁棒性好。
关键词：边缘提取；自适应；最大类间方差；阈值分割
中图分类号：TH162
文献标识码：A文章编号：1001-4551(2010)08-0010-05

LED wafer edge detection of an improved Canny algorithm
ZHI Shao-dan, LI Jian-yong, WANG Heng
（School of Mechanical and Electronic Control Engineering, Beijing Jiaotong University, Beijing 100044, China）
Abstract: In order to improve the automation level of the die bonders detection and the quality of extracted edge, an improved Canny edge-detecting algorithm was proposed based on Otsu for detecting the LED wafer edges. The proposed approach avoids artificial adjustment needed in Canny algorithm with fixed threshold value, and enhances the accommodation and robustness of the Canny edge-detecting algorithm. After the pixel value distributing analysis, maximum-variance between classes was used to obtain auto-adaptive threshold. And the value was made as the lower threshold one to detect the edge. The experiment results show that, this method is adapted to LED wafer edge-detecting. It can obtain more accurate wafer outline and poles edges and stay robust under the changed illumination.
Key words: edge detecting; auto-adaptive; Otsu; threshold