卡尔曼滤波磁光成像计盒维数焊缝跟踪算法

Seam tracking algorithm based on Kalman filtering of magneto-optical imaging using box-counting dimension

在激光焊接过程中精确地识别和跟踪焊缝十分重要。采用磁光成像传感器能够获取紧密对接微间隙焊缝(0~0.1 mm)的磁光图像。由于磁光图像存在较多干扰,根据计盒维数算法和卡尔曼滤波算法综合处理焊缝磁光图像,可获取较准确的焊缝位置。对焊缝磁光图像进行滤波去噪,将图像细分成块并计算出每个图像块的分形维数,再选取合适阈值分割图像,提取出焊缝位置参数。建立基于焊缝位置参数的系统状态方程和测量方程。应用卡尔曼滤波算法对焊缝位置进行最优状态估计,得到最小均方差条件下的焊缝偏差最优预测值。试验结果表明,卡尔曼滤波与计盒维数融合方法能够提高焊缝跟踪精度。

It is very important to accurately detect weld position and track the weld path in laser welding process.Magneto-optical sensor can be used to capture the magneto-optical images which contain the weld location information of micro-joint weld(0~0.1 mm).Since the magneto-optical images include much noises,the box-counting dimension algorithm and Kalman filter algorithm are used to deal with the magneto-optical images of the weld joint in order to capture the weld position accurately.The weld magneto-optical images are denoised and divided into blocks,and then the fractal dimension of each image block is calculated.An appropriate threshold is selected to segment the images,and the weld position parameters are extracted accurately.The state equations based on the weld position parameters and the measurement-equation for the weld position are established.A Kalman filter is developed to obtain the optimal prediction of the seam offset under the least squares condition through the optimal state estimation of weld position.Experiment results demonstrate that the accuracy of seam tracking can be improved significantly through combination of box-counting dimension algorithm and Kalman filter algorithm.