图像焊缝检测的圆形子窗口多步GPI方法  

Multi-step GPI method by circular sub-window for weld detection in image

作　　者：郑哲伟[1] 张文增[1] 陈强[1] 都东[1] 

机构地区：[1]清华大学先进成形制造教育部重点实验室,北京100084

出　　处：《焊接学报》2007年第8期77-80,共4页Transactions of The China Welding Institution

基　　金：国家自然科学基金资助项目(50275083);清华大学基础研究基金资助项目(JC2007009)

摘　　要：图像上紧密对接焊缝的提取是自动化焊接研究的难点。提出了焊缝检测新方法——圆形子窗口多步灰度投影积分(MSGPI)方法。在该方法中,利用圆形子窗口和灰度投影积分均值,克服了传统灰度投影积分(GPI)方法的缺陷;采用链表存储Sobel滤波所得边缘点,减少了计算量;用阈值法克服了划痕和锈迹干扰;利用划分角度区间分粗、精两步的检测措施,在保证精度的同时减少了计算量。结果表明,该方法能够快速准确地实现紧密对接焊缝的检测。It is significant but difficult to extract a weld in image when the gap of the weld is extremely small. This paper proposed a new method for weld detection, namely the multi-step gray projecting integral (MSGPI) method by circular sub-window. Firstly, the method applies a circular division in image as the sub-window for detection and the means of GPI values based on traditional gray projecting integral (GPI) method, which overcomes shortcoming of the latter one. Secondly, it adopts a link table to store edge points gained by Sobel filtering previously to decrease amount of computation. A threshold method is employed to avoid effectively interference caused by scratching and rust. Thirdly, the searching range of weld's angle by GPI process is divided into two stages: coarse and fine steps, which not only ensures detection accuracy but also greatly decreases amount of computation. The experimental results show that the MSGPI method can achieve rapid and accurate detection for weld even if the gap of weld is quite small.

关 键 词：图像处理 焊缝检测 灰度投影积分方法 焊缝跟踪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]