基于多视觉的熔池图像处理与特征提取  

Molten pool image processing and feature extraction based on multiple visions

作　　者：熊震宇[1] 顾网平[1] 王健[1] 薛诚[1] 

机构地区：[1]南昌航空大学航空制造工程学院,江西南昌330063

出　　处：《电焊机》2013年第5期93-96,共4页Electric Welding Machine

基　　金：江西省教育厅科技资助项目(GJJ11499);江西省教育厅科技资助项目(GJJ12448)

摘　　要：提出了一种多视觉焊接熔池采集系统,采用复合滤光的方法,能够有效地克服弧光的干扰,获取清晰的熔池图像。针对采集的熔池图像特点提出了一系列的图像处理方法。首先对熔池图像进行中值滤波去除图像中的噪声,然后采用CANNY多阈值方法提取熔池边缘特征,并利用去伪边缘、迭代拟合处理,从而获得熔池边缘特征。利用提取的熔池边缘特征和CCD摄像机与熔池的位置关系,得出熔池的后托角和堆高。将计算的熔池堆高与实测熔池堆高对比,偏差小于0.06 mm,证明了双视觉特征提取在熔池图像中的有效性和可靠性。A mohen pool image gathering system is designed based on multiple visions. It can effectively overeome the interference of the wehling arc,and obtain a clear molten pool image by using a composite filter. A series of image processing algorithms has been proposed based on the features of the molten pool image. A median filtering has been used to remove noise of the molten pool image. A CANNY muhi-threshold method has been used to extract the edge characteristics of the molten pool. Then the edge feature of the molten pool can be gotten after false edge removing and iterative fitting. The after-angle and height of the mohen pool can be extracted based on the position relationship of the CCD camera and the molten pool. Comparing the heap high of the molten pool between the calculation and the actual measurement,the deviation is less than 0.06mm： it proved the validity and reliability of the muhiple visions in the molten pool feature extraction.

关 键 词：多视觉 复合滤光 熔池图像处理 特征提取 

分 类 号：TG409[金属学及工艺—焊接]