激光焊接中窄焊缝特征解耦检测及多约束的视觉跟踪

The seam tracking has been a core problem for realizing the intelligence of robotic welding, which is a very important and valuable research topic.However,some scientic problems have still been open, such as feature detection and visual tracking of narrow seam for laser welding.Thus the project will study the follwing work under the background of the laser welding space curve narrow seam.First,narrow seam image feature extraction algorithm is designed. The phased coded kernel transformation method is proposed to focus the desired zone to a centerline, and then the robust and sub-pixel centerline can be obtained. Furthermore,the seam image feature point is gotten by employing the distance searching method and decoupling mapping principle of the designed visual system. Second,the repetive and non- Gaussian distribution characteristcs of the disturbance during the laser welding is analysized, then the minimum entropy filter design method is proposed by using the iterative frame and Renyi entropy performance index function.So,the optimal estimation of the seam image feature point can be solved. Third, the narrow seam tracking visual predictive tracking method is presented by using the nonlinear vision model and optimization design tool, which can drive the laser end to tracking the narrow seam in realtime. At the same time, these constraint conditions about vision field,seam feature bias, control input can be satisfied. The future results not only can be employed to solve the robust, decoupling, optimal recognization and visual tracking problems with multiple constraints of the narrow seam for laser weld, but also provide some useful basic theory knowledge for laser tracking the desired micro trajectory by using the visual system.

焊缝跟踪是焊接机器人智能化的核心问题，具有重要的研究意义。但在激光焊接中窄焊缝视觉跟踪理论研究上，如特征检测、视觉跟踪等方面有许多问题有待解决。为此，本项目以激光焊接空间曲线窄焊缝为背景，开展如下工作：1）窄焊缝图像特征解耦提取算法，提出基于相位编码核变换的方法，实现亚像素精度的中心线提取，利用基于距离搜索和视觉系统的解耦映射原理获取图像特征点；2）根据焊接环境干扰的重复性特征和非高斯分布特性，提出基于迭代学习和Renyi熵性能指标的滤波器设计方法，以解决图像特征点的最优估计问题；3）考虑视场约束、焊缝特征偏移量约束、控制量约束等条件，提出基于非线性视觉模型的窄焊缝跟踪视觉预测控制方法，实现激光束末端实时跟踪窄焊缝。本项目的预期成果不仅有助于解决窄焊缝的鲁棒、解耦、优化识别与多约束下的视觉跟踪等关键问题，也为机器人利用视觉系统识别微小目标，实现激光光束跟踪目标轨迹提供必要的理论基础。

目前，高能激光束对准窄焊缝主要利用昂贵的电气和机械定位系统来保证，但是这种方法不仅成本高昂，而且对焊缝环境变化的适应性差，不适用于复杂形状焊缝，难以处理焊接环境中出现的各种误差和变形的问题。尤其是，激光焊接对于焊缝的位置误差要求极高。对于聚焦后的斑点直径约为0.4毫米的激光而言，必须保证激光不从焊缝间隙中穿过去，因此，焊缝间隙中心线与激光束斑点中心的位置偏差一般要小于0.2毫米才能够满足激光焊接的要求。为了解决这类问题，焊接领域著名专家林尚扬院士曾指出：“视觉传感和控制技术是三项焊接核心技术之一，对提高焊接系统的自动化、智能化水平意义重大”。.为此，本项目的主要研究内容如下:设计了一种解耦视觉系统，实现激光焊接过程中的窄焊缝检测、识别与特征提取，为窄焊缝视觉跟踪控制系统提供焊缝状态变化的实时反馈信息，并考虑相机视场约束、视觉特征偏差约束等多个约束条件，在视觉空间解决激光焊接窄焊缝的高精度跟踪的优化控制问题。.在项目进行过程中，发表了SCI期刊论文三篇，EI期刊论文 篇，授权发明专利5项，授权实用新型专利5项，获得北京市科技进步三等奖1项，河北省科技进步三等奖1项，完成了规定的研究内容，解决了相关的关键问题。.本项目解决的窄焊缝视觉特征检测和跟踪控制问题不仅是激光焊接窄焊缝智能化的共性问题，也是先进制造等众多领域中，识别微小目标，进而实现激光光束跟踪目标轨迹方面，亟待解决的重要问题，具有十分重要的意义和价值。