针对激光拼焊焊缝边界特征点识别容易受噪声干扰, 识别鲁棒性不强问题, 提出了一种基于迭代的特征点识别方法。首先, 采用最大类间方差法与灰度重心法提取结构光条纹中心线; 其次, 结合条纹中心线畸变特征, 构造直线并计算中心点到直线距离, 获得距离曲线, 由距离曲线局部极值点对特征点进行初步定位; 再次, 在初步定位点的小范围内重新构造直线并进行迭代处理准确识别焊缝左右边界特征点; 最后, 通过序列图像信息对识别的特征点进行校正。采集不同焊接工艺参数下的激光拼焊焊缝图像进行实验, 左右边界特征点列坐标绝对误差均值与均方根误差均在2～4个像素。结果表明, 该方法识别精度高, 并具有较强的鲁棒性。

Aiming at the problem that the feature points recognition of weld seam boundary in tailored blank laser welding is easy to be disturbed by noise and is of low robustness, a feature points recognition method based on iteration is proposed. Firstly, the center line of the structured light stripe is extracted by the methods of the Otsu and grayscale barycenter. Secondly, a straight line is constructed by combing with the distortion of the stripe center line. And then the distances from the center points to the straight line are calculated and the distance curve is obtained. In addition, the feature points are preliminarily identified by the local extreme points of the distance curve. Thirdly, the straight line is reconstructed in the small area of the initial positioning point and use iterative processing to recognize the left and right boundary feature points of the weld seam accurately. Finally, the identified feature points are corrected by the sequence image information. The weld seam images in tailored laser welding under different welding process parameters were captured and the recognition experiments were carried out. The absolute mean error and root mean square error of column coordinates of left and right boundary feature point are 2 to 4 pixels. The results show that the method has high recognition accuracy and strong robustness.