【摘要】 为了满足自动化工艺的需要和获取更为清晰的焊缝原始图像，提出了一种基于YOLOv5和结构光的焊缝特征点提取算法。首先利用激光视觉系统获取焊缝图像，使用改进后的YOLOv5算法进行预训练；然后通过Steger算法提取结构光中心线，使用Harris算法对中心线图像进行处理，并确定焊缝坐标点的坐标；最后将坐标信息传输给PLC，控制十字滑台带动焊枪进行焊缝的跟踪操作。实验结果表明，改进后的YOLOv5算法mAP值更大、稳定性更高、鲁棒性更强。更多还原

【Abstract】 In order to meet the needs of automatic process and obtain more clear original images of welds, a feature point extraction algorithm based on YOLOv5 and structured light was proposed. Firstly weld images were obtained by laser vision system and pre-trained by the improved YOLOv5 algorithm.Then Steger algorithm was used to extract the structured light centerline, Harris algorithm was used to process the centerline image, and the coordinates of weld coordinate points were determined. Finally coordinate information was transmitted to PLC, and controled the cross slide table to drive the welding torch to track the welding seam. The experimental results show that the improved YOLOv5 algorithm has larger mAP value, higher stability and stronger robustness.更多还原