三维焊缝解耦视觉检测算法研究

发布时间：2018-08-19 18:21

【摘要】：随着现代化焊接生产对产品质量和生产效率的要求越来越高，如何保证焊接质量，提高焊接自动化水平成为摆在人们面前的一项重大课题。激光结构光主动视觉检测系统在提高焊接机器人的自动化、智能化方面发挥了积极作用。目前，国内外制造领域正面临着机器人大规模应用的宝贵机遇,对于能够提高机器人柔性和智能程度的视觉传感系统的需求日益迫切。因此，研究视觉检测算法，开发具有自主知识产权的视觉检测系统，对于提高机器人的感知能力具有十分重要的意义。以管道对接V形坡口焊缝为主要研究对象，设计了用于算法测试的视觉检测系统样机。选用基于Texas Instrument C64系列DSP处理器的智能摄像机、波长650nm的线激光光源、合适焦距的镜头、滤光片，并根据视觉检测系统的空间约束关系和视觉测量原理，设计了样机的硬件结构。通过对实时焊缝图像进行最小感兴趣区域(ROI)确定，自适应图像分割，中心线条纹提取，中心线差值与滤波，Hough变换提取结构光主直线等处理，采用基于距离搜索的方法提取焊缝特征点，准确获得了图像平面上反映焊缝形状和位置的特征点坐标。针对环形焊缝提出了解耦检测算法，即通过对单帧图像提取的三维焊缝特征信息进行解耦，计算出焊枪在二维平面内相对焊缝中心的偏差量。进而设计了基于图像的控制方法，实现了焊枪的快速准确定位。通过大量的现场测试结果证明，本文建立的视觉检测系统紧凑、可靠，设计的图像处理算法具有较好的稳定性和实时性、较高的可靠性和抗干扰性，，解决了焊接过程中强干扰环境中的焊缝识别和实时控制问题，能满足工业生产要求，具有重要的科学和使用价值。
[Abstract]:With the increasing demands of modern welding production on product quality and production efficiency, how to ensure welding quality and improve the level of welding automation has become an important issue in front of people. Laser structured light active vision detection system plays an active role in improving the automation and intelligence of welding robot. At present, the domestic and foreign manufacturing field is facing the precious opportunity of large-scale application of robot, and the need of vision sensing system which can improve the flexibility and intelligence of robot is increasingly urgent. Therefore, the research of vision detection algorithm and the development of vision detection system with independent intellectual property rights are of great significance to improve the perception ability of robots. The visual inspection system prototype for algorithm testing is designed, which takes V-groove weld seam of pipeline docking as the main research object. The intelligent camera based on Texas Instrument C64 series DSP processor, the line laser light source of wavelength 650nm, the lens with suitable focal length and the filter are selected. The hardware structure of the prototype is designed according to the spatial constraint relation of the vision detection system and the principle of vision measurement. In this paper, the minimum region of interest (ROI), adaptive image segmentation, centerline stripe extraction, centerline difference and filter Hough transform are used to determine the minimum region of interest of the real-time weld image. The distance search method is used to extract the weld seam feature points, and the coordinates of the feature points reflecting the shape and position of the weld seam on the image plane are obtained accurately. A decoupling detection algorithm is proposed for annular welds, that is, by decoupling 3D weld characteristic information extracted from single frame image, the deviation of welding torch relative to weld center in two-dimensional plane is calculated. Furthermore, the control method based on image is designed to realize the fast and accurate positioning of welding torch. Through a lot of field test results, it is proved that the visual inspection system established in this paper is compact, reliable, and the image processing algorithm designed has good stability and real-time, high reliability and anti-interference. The problem of welding seam identification and real-time control in the environment of strong interference during welding process is solved. It can meet the requirements of industrial production and has important scientific and practical value.
【学位授予单位】：河北工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.41

【参考文献】