SIA OpenIR  > 智能产线与系统研究室
基于视觉的机器人激光焊接质量检测技术研究
Alternative TitleResearch on Robotic Laser Welding Inspection Based on MachineVision
张雷1,2
Department现代装备研究室
Thesis Advisor赵明扬
ClassificationTP242
Keyword激光拼焊 立体视觉 图像处理 摄像机标定 空间规划
Call NumberTP242/Z33/2008
Pages67页
Degree Discipline机械电子工程
Degree Name博士
2008-05-30
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract视觉检测技术是随着计算机视觉技术和光电技术的飞速发展,而出现的一种新的检测技术。检测被测目标时,把图像当作检测和传递信息的手段或载体,从图像中提取有用的信号,它是以现代光学为基础,融光电子学、计算机图像学、信息处理、计算机视觉等科学技术为一体的现代检测技术。现代激光自动化焊接技术是由激光、计算机、机器人、数控和精密机床等相结合的综合高新技术,此项技术已成为工业生产自动化的关键技术,拥有普通加工技术所不能比拟的优势。为了克服机器人焊接过程中各种不确定因素对焊接质量的影响,提高机器人作业的智能化水平和工作可靠性,要求焊接机器人系统不仅能实现空间焊缝的实时跟踪,而且还能实现焊接参数的在线调整和焊缝质量的实时控制,即焊接机器人焊接过程的自主化和智能化。本文的研究依托于中国科学院知识创新工程方向性项目“全自动激光拼焊成套装备生产线”,旨在探索立体视觉检测系统的实现及其在激光拼焊工程中的应用的问题。从理论和实践两个方面,对其中的若干关键技术,如视觉检测系统创新设计、数学模型、量化误差、摄像机标定、结构光条纹中心线提取、焊前特征检测、溶池边缘提取、焊后缺陷图像匹配算法、三维重建和表面孔的视觉定位等进行了研究。主要研究成果如下: 1.提出了一种可以用于焊前跟踪,焊后检测,以及焊接过程中对激光溶池进行监测的多功能激光视觉检测装置。推导了检测系统在不考虑像平面安装倾斜角度时和考虑像平面安装倾斜角度时检测点坐标的计算公式以及量化误差公式,分别针对由于数模转换量化误差、安装角度倾斜误差、安装高度误差三个方面引起的量化误差,分析其关于行,列,以及不同倾斜角度的影响分布规律。并对于各种情况进行了仿真,对于各种误差分布特征进行了分析,提出了检测奇异点的情况和数学模型的局限性。以上工作为实现焊缝三维信息的高精度提取奠定了基础。 2.对于摄像机的标定技术进行了研究,结合工程实际,利用zhang的标定法和matlab标定工具包,对于摄像机进行了标定;针对检测相机视场较小,标定采集范围不易调整和相机的畸变主要发生在视场边缘等特征,在保证要求的精度范围之内,提出一种基于标定靶的标定方法,实验证明该方法的标定与测量精度能够满足工程需要。 3.研究了现有的条纹中心提取算法过程,提出了基于OTSU阈值的多次高斯拟合平均法和基于OTSU阈值的质心平均法计算激光条纹中心坐标。该法对条纹的噪声,散斑和被测工件表面漫反射有很强的抵制作用,因此具有很强的鲁棒性。实验表明,与传统方法相比,具有更高的提取精度。同时为了适应激光条纹被工件表面调制后发生的角度变化,以及硬件安装带来的激光线型条纹倾斜,提出了一种自适应方向模板法,可以解决特殊倾斜角度时的激光条纹中心线提取问题。三个仿真试验验证了方法的可行性。 4.提出了一套在线实时进行焊前检测的图形处理算法,可以实现焊缝宽度,焊前错配和焊缝中心位置检测。通过工程实验提取了各指标的检测结果,并验证了算法的正确性。 5.提出了一种基于数学形态学的激光拼焊溶池边缘检测算法,对于激光拼焊中的溶池图像进行边缘提取,基于真实图像进行了实验研究表明,提取边缘效果可以达到单像素。 6.对焊后表面形貌检测的图像实时处理算法进行了研究。提出了一套在线实时进行焊后焊缝表面缺陷检测的图形处理算法,可以进行焊缝宽度,错配,凹度,凸度,咬边,焊接倾角,过高七种表面焊接缺陷的匹配识别;对于整个焊缝的表面形貌进行三维重建。通过等厚板焊接和不等厚板焊接两种试验,验证了算法的合理性和鲁棒性。提出了一种基于randon变换的错配和兴趣区域快速检测算法。对于表面孔的检测算法进行了探讨,主要针对孔的检测中的噪声和表面反射,研究了腐蚀膨胀对于表面孔定位和大小的检测影响。 7.研究了以普通6R机器人进行焊缝视觉检测的工作视野,即焊缝视觉检测空间,提出了一种生成焊缝视觉检测空间的解算法;以此检测视野为依托,探索了在视觉检测视野中进行焊缝视觉检测的初始位置规划问题,提出了一种初始位置规划算法;仿真结果证明了算法的正确性。
Other AbstractWith the rapidly growth of the computer vision and photoelectricity,visual inspection technique is an emerging multi-discipliary field with many challenges. It combines different aspects of modern optics, computer iconography,information processing, computer vision, pattern recognition, automation, and control system. Laser weiding technlogy is a synthetic technology involved with plenty of subjects, it reflects the development degree of a country in science and high technology industry. Laser welding is a high and new technology based on techniques of laser, computer, robot, numerical control machine. It has been a key technique in industry’s production automation. In order to improve welding robot’s intelligence and robust, overcome bad influence to welding quality owing to many uncertain factors, Welding robot is essential to be examined by doing a 3D-piece run and measuring variability of the joint position and joint gap. Also, this information must be fed back upstream to improve the detail parts and the fixturing. The variability must be reduced so the total laser system process capability is sufficient to produce quality welds. The research in this paper is supported by Chinese Academy of Sciences directional project-“Research on the new automatic laser welding line for tailor-welded blanks”. The purpose of this research is to realize the stereo vision inspection system and solve the practical problems occurred between robot arm and stereo vision system when the system is used in project. Some of the critical techniques are investigated from theory and practice. The author pays more attention to the research on a creative design of robotic visual tracking system based on CMOS, mathematical model, 3-D measurement quantization error, camera calibration, the mid-line of the laser stripe extraction, pre-weld inspection,weld-pool edge inspection, post-weld inspection, defect image matching, 3D reconstruction and robot inspection localization plan based on stereo vision. The main productions of this paper are as follows: First, in order to realize autonomous moving and automatic weld seam tracking for welding robot in Tailed Welded Blanks, a creative design of robotic visual tracking system based on CMOS has been developed. The new system is not only used to inspect the worcckpiece ahead of a welding torch and measure the joint orientation and lateral deviation caused by curvature or discontinuites in the joint part, but also to record and measure the image size of the weld pool or plasma plume. Moreover, the hardware and software components are discussed in brief. The mathematical models and error analyses of visual inspection system with horizontal image plane and with sloping one are proposed. The quantization errors about A-D quantization, adjustment angle and adjustment altitude is provided. Their distributed functions about line, column and angle are analysed. Simulate process of the analysis is adopted to test the correct of the analysis.Inspection oddness point and limitation of the model are also put forward. These works are the basis of realizing high accurate inspection. Camera calibration is an important step for 3D inspection and its precision decide the later precision for inspection. Popular camera calibration methods are analysed in the second chapter. In practice the nonlinear method of Zhang and the Camera Calibration Toolbox for Matlab have been applied well in the system. A novel calibrate method based target is put forward which fits well whithin the small field of view and few collecting specimen in this project. The experiment results prove that this method can meet improve the accuracy of inspection. Third, Present methods how to extract the center of the laser stripe are dicussed. A k-times Gauss average method based on OTSU threshold and a centroidal average method based on OTSU threshold to draw the center of the laser stripe are put forward. New methods. As a result, the center of the laser stripe image can be located more accurately. Because the new method can resist noise,speckle,and the workpiece sufarce scattered reflection. Comparing with the traditional method, the accuracy of our method are better. New method for obtaining the center of structured light stripe by self adapting template is put forward in order to adapt to the sloping angle of the laser stripe which made by hardware and adjustment angle. Three simulate experiments have been desigened to illustrate the method’s feasibility. Fourth, A pre-weld inspection method is proposed to track weld seam and inspect mismatch and gap width on line. The experiment results show that these methods can inspect the named defects. Fifth, a new flowchart model for image processing of weld pool is proposed using the principles based on the inspection and monitoring system in Tailored Welded Blanks. Furthermore, three imperceptibility edges are detected well in the experiments. The experimental results show that the new method reaches not only perfect edge detection result but also good robustness to noise in TWB. After compared with other well-known edge detection algorithms.This technology performs favorably and can be used widely in TWB. Sixth, A novel profile inspection algorithm of tailored welded blanks about on-line image process and features extraction for structured light image of laser welding has been developed.The method can inspect beadwidth,mismatch, convexity and concavity, undercut, sloping angle, and overthickness defects. 3D reconstruction of the whole welding seam also is given. Many experiments are carried out to examine the effect of the algorithm. From the experiment results, it can be said that the algorithm proposed in this paper is feasible on principle and has directive meaning for realization of on—line inspection of Tailored welded blanks. In order to inspect the mismatch and set the range of interest, the Radon Transform is applied to determine the x and y coordinate in image coordinate system. Analyze how to inspect the hole on the surface of the blanks,especially, erosion and dilation are used to inspect the hole location and diametral preventing the disturbance of noise and reflection. Lastly, the calculation method of visual workspace of a 6R spatial manipulators is given. on the basis of visual workspace,a initial point planning algorithm for the inspected welding seam is proposed. Simulation instances of a 6R manipulators demonstrate the validity of this proposed algorithm.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/404
Collection智能产线与系统研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
张雷. 基于视觉的机器人激光焊接质量检测技术研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2008.
Files in This Item:
File Name/Size DocType Version Access License
LWZL001.pdf(2389KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张雷]'s Articles
Baidu academic
Similar articles in Baidu academic
[张雷]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张雷]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.