基于最小二乘多项式的焊缝图像拟合研究

Weld extraction based on least square polynomial fitting method

针对不同类型焊缝线提取拟合精度低的问题,提出基于最小二乘多项式的拟合方法。利用实验跟踪系统,将采集后的焊缝图像进行中值滤波和二值化,在二值化图像的基础上提取焊缝的位置坐标。建立拟合焊缝的多项式函数模型,根据最小二乘原理,得到多项式函数最优估计解。Matlab仿真实验结果表明,九阶拟合函数的决定系数达到99.94%,最大残差为5.62 mm,在精度范围内能够准确拟合焊缝图像。

In order to solve the problem of low precision of weld extraction,a fitting method based on least square polynomial was proposed.An experiment racking system was used to filter and binarize the image after the weld was collected by the visual sensor.Based on the binarization image,the location information of the weld was extracted,and a polynomial function model was established.According to the principle ofleast square,the optimal estimation solution of the polynomial function was obtained.The polynomial order and fitting results were determined by the coefficient of determination(r⁃squar),standard deviation and residual.The fitting results showed that the determination coefficient of the ninth order fitting function reached 99.94%and the maximum residual was 5.62 mm,which could accurately fit the weld image.