基于Mask R-CNN和结构光的焊缝特征信息检测方法研究

Research on Weld Feature Information Detection Method Based on Mask R-CNN and Structured Light

为了提高焊缝跟踪系统的精准度,提出了一种基于Mask R-CNN和结构光的焊缝特征参数检测系统。通过Mask R-CNN对焊缝结构光图像提取线结构光边界信息,再通过灰度重心法完成对结构光中心线的提取;对中心线图像进行斜率分析,来确定焊缝特征点坐标;最后将坐标信息传输至PLC中,通过PLC控制十字滑台带动焊枪进行焊缝的跟踪操作。避免了强光、灰尘和噪声等干扰,提高了焊缝特征信息检测的准确性。通过与Faster R-CNN和YOLOv5的预测实验对比,Mask R-CNN在焊缝检测中的准确度和稳定性更高。

In order to improve the accuracy of the weld tracking system,a weld feature parameter detection system based on Mask R-CNN and structured light was proposed.The Mask R-CNN was used to extract line structured light boundary information from the weld structured light image,and then the gray center of gravity method was used to extract the centerline of the structured light.The slope analysis was performed on the centerline image to determine the coordinates of the weld feature points.Finally,the coordinate information was transmitted to the PLC,and controlled the cross slide table through PLC to drive the welding gun for weld tracking operation.It avoids interference such as strong light,dust and noise,and improves the accuracy of weld feature information detection.Through comparison with Faster R-CNN and YOLOv5 in prediction experiments,Mask R-CNN has higher accuracy and stability in weld detection.