针对自适应性低的焊缝跟踪系统在实际焊接环境中易受噪声干扰的问题,结合深度卷积神经网络强大的特征表达能力和自学习功能,研究了基于深度分层特征的焊缝检测和跟踪系统,该系统可精确地从噪声污染的时序图像中确定焊缝位置。为彻底解决焊枪依循计算轨迹运动所出现的抖振问题,设计了模糊免疫自适应的智能跟踪控制算法。实验结果显示,在强烈弧光和飞溅的干扰下,传感器测量频率达20Hz,焊缝跟踪精度约为0.2060mm,且焊接过程中焊枪末端运行平稳。该系统能实现焊缝平滑的实时跟踪,抗干扰能力强,焊缝轨迹跟踪准确,能满足焊接应用要求。 In view of the problem of low adaptability of weld seam tracking system in the actual welding environment, combined with the strong feature expression ability and self-learning function of deep convolutional neural network, the weld seam detection based on depth stratification and Tracking system that accurately determines seam locations from noise-contaminated time-series images. In order to solve the chattering problem that the welding torch appeared in tracking motion, a fuzzy immune adaptive intelligent tracking control algorithm was designed. The experimental results show that under the interference of intense arc and spatter, the measuring frequency of the sensor is up to 20Hz, the seam tracking accuracy is about 0.2060mm, and the end of the welding torch runs smoothly. The system can achieve smooth and real-time tracking weld, anti-interference ability, accurate tracking of weld trajectory, to meet the welding requirements.