焊缝结构光图像处理与特征提取方法

Image Processing and Feature Extraction for Structured Light Images of Welded Seam

为实现焊缝磨抛自动化,将双CCD相机与激光器搭载于磨抛机器人上,构成机器人视觉导航和检测系统。根据机器人与焊缝特征点之间空间位置的相关性,构造动态感兴趣区域(ROI),提出了一种提取焊后焊缝结构光特征线的快速算法,该算法可将图像处理区域缩小到原来的1.49%,从而极大地提高了运算效率。在动态ROI内,对图像进行分析,针对焊缝结构光图像的特点,对图像的预处理过程进行了优化,采用动态高斯平滑模板处理直方图,改进了阈值计算方法。在此基础上,提取焊缝激光带的特征线,并进行了实验研究。研究结果表明,视觉系统稳定可靠,提出的算法能够快速地锁定动态ROI,可准确地提取结构光特征线,从而为磨抛机器人视觉导航和检测奠定了基础。

To automatically grind and polish welded seam,a binocular stereo is constructed by connecting double CCD and laser with grinding and polishing robot.On the dependency between the robot and structured light on welded seam,a dynamic region of interest（ROI） is composed,and an algorithm for quick-extracting typical line is proposed as well.This proposed algorithm enables to narrow the image processing region to 1.49% of the original one,to significantly improve the efficiency.Analyzing the image within dynamic ROI,the preprocessing is optimized according to the characteristic of quick-extracting image,and the threshold computing during binarization also improved by histogram smoothed by a dynamic Gaussian mask.Then the typical lines of structured light band are extracted and experimentally investigated.It indicates that the visual system is endowed with good reliability and stability,and the algorithm is able to quickly locate dynamic ROI and precisely extract laser featured line.