Abstract: To solve the problem of limited space and disturbance of seam tracking accuracy when welding large work pieces,an embedded visual tracking control system for autonomous mobile welding robot was proposed and designed,which was suitable for field welding operation with its compact and small size.A seam tracking algorithm based on Kernelized correlation filter was proposed to solve the problem of arc interference during welding and to realize real-time,accurate and reliable seam tracking.A large number of positive and negative samples were trained to construct classifiers and Gaussian kernel function mapping was used to improve tracking accuracy and reliability.Sample cyclic matrix and Fourier transform were calculated to reduce the amount of calculation and realize the real-time tracking.The automatic tracking algorithm was implemented in the embedded system based on ARM to control the crawling and yaw of the mobile robot and realize real-time deviation correction in welding process.The welding experiment of V-shaped welding seam for large pipeline was conducted,and the frame rate of the visual sensor could reach10fps.The proposed system could correct the deviation automatically with the movement of the welding robot,the track of the weld was accurate,and the welding surface was smooth.

Key words: autonomous welding robot, embedded vision control, seam tracking, kernelized correlation filter, rectify deviation control

摘要： 为解决大型工件现场焊接空间受限和焊缝跟踪精度受干扰的问题,提出并设计了一种自主移动焊接机器人嵌入式视觉跟踪控制系统。系统紧凑、体积小,适应现场焊接作业。提出基于核相关滤波的焊缝跟踪算法,以解决焊接时的弧光干扰问题,实现实时、精确、可靠的焊缝跟踪。该算法通过生成大量的正负样本进行分类器训练,并映射到高斯核函数空间进行计算,来提高跟踪精度和可靠性;通过构建样本的循环矩阵和傅里叶变换来减少计算量,提高跟踪实时性。在ARM嵌入式系统中实施该自动跟踪算法,完成了对移动机器人的爬行和横摆控制,并实现了焊接过程的实时纠偏。通过对大型管道V型焊缝进行焊接实验表明,帧率达到10 fps,焊缝轨迹跟踪准确,焊接表面平顺。

关键词: 自主移动焊接机器人, 嵌入式视觉控制, 焊缝跟踪, 核相关滤波, 纠偏控制