基于改进KCF和峰值跟踪的焊缝识别算法

Weld Seam Identification Algorithm Based on Improved KCF and Peak Tracking

针对焊接中产生的烟尘和弧光等干扰导致的焊缝跟踪困难问题,提出了一种基于改进核相关滤波算法(kernelized correlation filters, KCF)的焊缝跟踪方法。首先,引入灰度特征使得改进KCF算法对焊接烟尘干扰有更强的抗扰能力;然后,提出了一种基于二次指数平滑法的峰值跟踪算法对实时的焊缝位置进行预测跟踪;最后,借助Hamming窗和余弦相似度对焊缝特征点所在的位置进行补偿。实验结果表明,该方法在焊接烟尘和弧光的干扰下具有较高的跟踪精度,且同样适用于多种焊接任务,具有较强的适应性。

Aiming at the difficulty of seam tracking caused by interference such as smoke and arc light in welding,a seam tracking method based on improved kernelized correlation filters(KCF)is proposed.Firstly,the gray feature is introduced to make the improved KCF algorithm more robust to welding smoke interference.Then,a peak tracking algorithm based on quadratic exponential smoothing method is proposed to predict and track the real-time weld position.Finally,the location of weld feature points is compensated by Hamming window and cosine similarity.The experimental results show that this method has high tracking accuracy under the interference of welding smoke and arc light,and is also suitable for various welding tasks with strong adaptability.