一种激光视觉引导的自动识别V形焊缝的算法

An Algorithm for Automatic Recognition of V Shape Weld Guided by Laser Vision

如何准确地识别线结构光在V形焊缝上留下的光带是机器视觉的一个关键技术,对此文章提出一种新的图像处理方法。首先对原图像进行中值滤波。然后选用一种新的高斯模板对其进行滤波,可以有效地去除面积较大,宽度较宽的噪声。接着采用一种新的ROI区域提取算法,可以有效地确定光带所在位置,提高运算效率。然后采取孤点滤波删除散落在图像上的曝光点。接着对其进行细化,提取激光条纹中心线。提取焊缝特征点时,运用斜率法将图像中的点进行分类随后进行最小二乘法的直线拟合,得到特征点坐标。然后将特征点坐标从图像坐标系转到基坐标系下。最后通过计算得到焊缝参数。实验结果表明:V形坡口的角度、高度、宽度的计算结果与实际值的相对误差分别为0.33%、0.585%、0.186%。满足的抗干扰能力强、精度高和实际焊接等要求。

How to accurately identify the light band that line structured light left in a V-shaped weld is a key technology of robot vision,this paper proposed a new method of image processing. First,the original image is filtered by median filter. Then use LOG filtering with a new template which can eliminate the large area and wide width noise. And next,extract the ROI region by use a new algorithm which can determine the location of the light band and improve the operation efficiency. Then delete the exposure point scattered on the image by particle filtering. After that,extract centerlines of laser stripes by thinning algorithm. In order to extract the feature points,the paper take the slope method to classify the points in the image,then the least square method is used for the linear fitting and got the welding parameters. Then,the feature point coordinates are transferred from the image coordinate system to the base coordinate system. Finally,weld parameters are obtained by calculating. Experimental results indicate that the relative error of V groove angle,height,width between the calculation result and the actual value are 0. 33%,0. 585% and 0. 186%,respectively. It can satisfy the requirements of the strong anti-interference ability,high precision and actual welding.