图像分割在主动光焊缝检测中的应用  

APPLICATION OF IMAGE SEGMENTATION ON WELD LOCATION USING ACTIVE LASER VISION SENSOR

作　　者：徐培全[1] 芦凤桂[1] 鲁建波[1] 姚舜[1] 

机构地区：[1]上海交通大学,200000

出　　处：《焊接》2006年第3期31-35,共5页Welding & Joining

基　　金：国际科技合作重点基金资助项目(2004DFA02400);上海市科学技术委员会科研计划基金资助项目(036105001)

摘　　要：开发了一种基于环形激光的视觉传感器并用于机器人焊接中的焊缝定位与跟踪,重点研究了在无弧光和弧光条件下焊缝图像滤波与图像分割等技术,解决了传感器图像采集与处理所面临的问题。在分析了最大类间方差自动取阈法、模糊簇聚方法的基础上,提出了一种基于灰度最大值统计与分区的两步法自适应阈值分割技术。结果表明,在无弧光条件下,利用灰度最大值统计方法获得了满意的分割效果,结合灰度最大值统计与区域分割技术适用于弧光条件下焊缝图像的分割,这两种自适应分割方法简单方便,优于传统的最大类间方差自动取阈法与模糊簇聚方法。此外,通过顺序统计量滤波技术很好地解决了光线弱化问题。Vision sensor based on circular laser trajectory used for weld location and seam tracking was developed. Image filter and segmentation technique was studied under GTAW （Gas tungsten arc welding） arc condition or without welding arc condition, and problems of image acquisition were solved. On the basis of Otsu and fuzzy c-means （FCM） analysis, we presented two-step segmentation （TSS） technique. The results showed： under no arc condition, maximum gray level method obtained complimentary effect. Satisfactory segmentation results were obtained using two - step image segmentation technique. Combined with partition allocation methods, weld image under GTAW arc condition was segmented successfully. Two image segmentation methods are convenient and superior to Otsu an FCM method. In addition, proposed order - statistic filtering technique is valid for circular laser trajectory perfection.

关 键 词：环形激光 图像分割 焊缝图像 

分 类 号：TG441.7[金属学及工艺—焊接] TP391.41[自动化与计算机技术—计算机应用技术]