基于线结构光的亚像素精度焊缝提取方法研究

对于激光视觉焊缝跟踪系统,基于线性结构光快速、高精度地提取焊缝特征点是系统搭建的关键。现有算法多是采取像素级别的提取特征,现提出改进的亚像素精度算法用以提取焊缝特征点。与以往算法不同的是,算法不需要进行阈值的选取,提取条纹中心线和检测特征点的过程,都采用了先计算出亚像素位置,再对图像进行处理,显著地提高了算法的精度。并且目前图像处理多采用深度学习,但都为对像素的离散点实现,难以做到亚像素精度。实验结果表明,该算法能够满足生产实际要求,能够实时、精确地实现焊缝提取。

Sub-pixel precision seam extraction method based on line structured light

For the laser vision seam tracking system, it is the key to extract the weld feature points quickly and accurately based on the linear structured light. Most of the existing algorithms are pixel-level feature extraction. Now an improved sub-pixel accuracy algorithm is proposed to extract weld feature points. Different from the previous algorithm, the algorithm does not need to select the threshold value. In the process of extracting the fringe centerline and detecting the feature points, the sub-pixel position was calculated first, and then the image was processed. The accuracy of the algorithm was greatly improved. Moreover, image processing mostly adopts deep learning technology at present. But the deep learning technology is all for the realization of pixel discrete point, it is difficult to achieve sub-pixel precision. The experimental results showed that the algorithm could meet the actual production requirements, and could achieve real-time and accurate weld extraction.