CASIA OpenIR  > 复杂系统管理与控制国家重点实验室  > 先进机器人
基于结构光视觉的焊缝识别与跟踪技术研究
范俊峰
Subtype博士
Thesis Advisor景奉水
2019-05
Degree Grantor中国科学院大学
Place of Conferral中国科学院自动化研究所
Degree Discipline控制理论与控制工程
Keyword初始点导引 焊缝跟踪 焊缝类型识别 焊缝特征提取 结构光视觉
Abstract

焊接作为一项重要的制造工艺,在工业生产中有着广泛的应用。当前大多数
焊接机器人属于“示教-再现” 和离线编程机器人, 它们有个明显的缺点: 只能
沿着预先示教或编程的轨迹进行焊接, 无法适应由于工件加工误差、装配误差、
焊接热变形等引起的焊缝位置的变化,降低了焊接质量。 为了使焊接机器人自主
感知焊接环境实现智能化焊接,需要给焊接机器人安装焊缝传感器。 由于结构光
视觉传感器具有非接触、信息量大、测量精度高、抗干扰能力强等优点,受到了
研究人员的青睐。 目前, 虽然已有一些结构光视觉传感器用于机器人焊接,但是
存在一系列的不足: 不能实现焊缝类型的自动识别、焊缝特征点提取方法计算量
大灵活性差、不能适用于微细焊缝的初始点导引和跟踪等,因此研究基于结构光
视觉的焊缝识别与跟踪技术具有重要的理论意义和工程价值。
本文针对基于结构光视觉的焊缝识别与跟踪技术展开研究,内容主要包括结
构光视觉传感器的设计与标定、焊缝类型的识别、焊缝特征点提取、微细焊缝初
始点导引和跟踪。 论文的主要工作如下:
1. 针对机器人智能化焊接的需要,设计了基于结构光视觉的焊缝跟踪传感
器并提出了一种便于机器人现场应用的结构光平面参数标定方法。本文设计的传
感器有两个主要的特点:一是采用多层抽屉式减光滤光系统,提高传感器的环境
适应能力;二是采用模块化结构,便于功能扩展和维护。 本文提出的结构光平面
参数标定方法,通过控制机器人末端有约束地运动一次,在标定靶标上产生两条
不共线的激光条纹,完成结构光平面参数的标定。 该方法不需要加工高精度的三
维标定靶标,也不需要手工测量,标定过程简单实用。
2. 针对现有的焊缝类型识别方法计算量大识别率低的缺点,提出了一种基
于 SVM 的常见焊缝类型的识别方法,该方法可以快速准确的识别出焊缝类型。 首先对采集到的结构光图像进行处理得到分类模型的输入特征向量, 然后以得到的特征向量作为输入, 以其所属的焊缝类型作为输出, 通过训练建立 SVM 焊缝类型识别模型。最后针对常规的基于几何特征的焊缝特征点提取方法应用于焊缝跟踪过程中存在的计算量大灵活性差的缺点,提出了一种基于高效卷积算法的焊缝特征点提取方法。该方法可以适用于不同的焊缝类型如对接坡口焊缝、搭接焊缝等。即使在焊接噪声的干扰下,该方法也能够快速鲁棒的提取得到焊缝特征点。
3. 针对传统的结构光视觉传感器不适用于微细焊缝初始点导引以及基于被
动光视觉的焊缝初始点导引方法存在鲁棒性差的不足,设计了双源结构光视觉传
感器并提出了微细焊缝初始点导引方法。双源结构光视觉传感器用于获取包含激
光条纹和微细焊缝的高信噪比图像,完成微细焊缝的识别和测量。 在此基础上,
提出了基于两阶段的平面微细焊缝初始点导引方法和基于结构光扫描的曲面微
细焊缝初始点导引方法。 平面微细焊缝初始点导引方法具有导引精度高、导引速
度快的特点, 曲面微细焊缝初始点导引方法具有适用范围广的特点。
4. 基于所设计的双源结构光视觉传感器,提出了一种适用于平面和曲面微
细焊缝跟踪的方法,克服了传统的基于单源结构光视觉的焊缝跟踪方法不适用于
微细焊缝跟踪以及基于被动光视觉的微细焊缝跟踪方法图像处理复杂、鲁棒性差
的不足。 该方法可以同时获得焊枪水平竖直两个方向的偏差,并在两个方向上独
立设计相互解耦的控制器,利用误差滤波、模糊 PID 控制器和输出脉冲验证来实
现微细焊缝水平竖直方向同时准确跟踪。
最后, 对本文的研究工作进行总结,并提出了下一步的研究计划。
 

Other Abstract

As an important manufacturing process, welding is widely used in industrial
production. At present, most welding robots belong to teach-and-playback and
off-line programming robots. They have an obvious shortcoming. They can only
move along the recorded path and cannot adjust the welding torch position to adapt to
the change of the position of the workpiece caused by machining error, assembly error
and welding thermal deformation, which reduces welding quality. In order to make
the welding robot autonomously perceive the welding environment and realize
intelligent welding, it is necessary to install welding sensors for the welding robot.
Due to the advantages of non-contact, abundant information, high precision and
strong anti-interference ability, the structured light vision sensor is favored by
researchers. Currently, although some structured light vision sensors have been used
in robot welding, there exist a series of shortcomings. For example, the automatic
recognition of weld seam type is not realized. The seam feature extraction methods
lack flexibility and computational efficiency. Moreover, these sensors are not well
suited for initial point guiding and seam tracking of micro gap weld seam. Therefore,
it has significant theoretical meaning and engineering value to study seam recognition
and tracking technology based on structured light vision.
In this thesis, the seam recognition and tracking technology based on structured
light vision is studied including the design and calibration of the structured light
vision sensor, weld seam type recognition, seam feature point extraction, initial point
guiding and seam tracking of micro-gap weld. The main work of this thesis is as
follows:
Aiming at the need of robot intelligent welding, a seam tracking sensor based on
structured light vision is designed and a simple and practical calibration method of
structured light plane parameters is presented. The developed sensor has two main
characteristics: the multi-layer drawer type dimming filter system is designed to
improve the environmental adaptability of the sensor, and the modular structure
design method is adopted to facilitate functional expansion and maintenance. By
controlling the constrained movement of the robot's end at one time, two non collinear
laser stripes are generated on the calibration target to complete the calibration of the
structural light plane parameters. This method does not need high-precision 3D
calibration targets or manual measurement, and the calibration process is simple and
practical.
Aiming at the shortcomings of huge computation cost and low recognition rate of
the existing weld seam type recognition method, a weld seam type recognition method
based on support vector machine (SVM) is presented, which could achieve weld seam
type recognition quickly and accurately. Firstly, the captured structured light images
are processed to get the input feature vector of the classification model. Then, taking
the feature vector as input and the weld seam type as output, the SVM weld seam type
recognition model is established through training. Finally, aiming at the shortcoming
of the conventional seam feature point extraction method based on geometric features,
which requires a large amount of calculation and has poor flexibility in the process of
welding seam tracking, a seam feature point extraction method based on efficient
convolution algorithm is proposed. The method could be applied to different weld
types, such as butt groove weld and lap weld. Even under the interference of welding
noise, the method could extract the seam feature points quickly and robustly.
Aiming at the problem that traditional structured light vision sensor is not
applicable to initial point guiding of micro-gap weld and initial point guiding method
based on passive vision sensor has disadvantages of poor robustness and low guiding
accuracy, a dual-source structured light vision sensor is developed and initial point
guiding methods of micro-gap weld are proposed. The dual-source structured light
vision sensor is used to capture the high SNR image including laser stripe and
micro-gap weld seam, and achieve the recognition and measurement of micro-gap
weld. On this basis, initial point guiding methods of planar micro-gap welds based on
two stages and initial point guiding methods of cruve micro-gap welds based on
structured light scanning are presented. The initial point guiding method of planar
micro-gap weld has the characteristics of high precision and fast guiding speed. The
initial point guiding method of curved surface micro-gap weld has the characteristics
of wide application.
Based on the designed dual source structured light vision sensor, a seam tracking
method suitable for planar and curved micro-gap weld is proposed, which overcomes
the problems that traditional seam tracking method based on single source structured
light vision is not suitable for micro-gap weld and seam tracking method based on
passive vision has disadvantages of complex image processing and poor robustness.

This method could obtain the welding torch deviations in both horizontal and vertical
deviations, and independently design the decoupling controller in both directions. The
error filtering, Fuzzy-PID controller and output pulse verification are used to realize
the accurate tracking of micro-gap weld seam in both horizontal and vertical
directions.
Finally, the research work of this thesis is summarized and the future research
plan is proposed

 

Pages150
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/23923
Collection复杂系统管理与控制国家重点实验室_先进机器人
Corresponding Author范俊峰
Recommended Citation
GB/T 7714
范俊峰. 基于结构光视觉的焊缝识别与跟踪技术研究[D]. 中国科学院自动化研究所. 中国科学院大学,2019.
Files in This Item:
File Name/Size DocType Version Access License
基于结构光视觉的焊缝识别与跟踪技术研究.(6567KB)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
[范俊峰]'s Articles
Baidu academic
Similar articles in Baidu academic
[范俊峰]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[范俊峰]'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.