论文部分内容阅读
机器视觉系统对于提高激光机器人再制造质量有重要作用。针对浅斑类缺陷无法在点云数据中定位的问题,开发了专用识别系统,将二维图片和三维点云结合起来实现该类缺陷的三维检测。对灰度图片进行处理,包括图像频域增强、区域标记和区域合并等,实现了二维图片中的缺陷定位;利用双目立体视觉系统对再制造零件表面进行扫描,获取零件表面三维点云数据,同时将二维缺陷边界转换为三维缺陷边界,实现了点云数据中缺陷的定位。实验结果表明,该系统能有效识别浅斑类缺陷,并且再制造精度高。
Machine vision system plays an important role in improving the remanufacturing quality of laser robots. Aiming at the problem that shallow spot defects can not be located in point cloud data, a special recognition system has been developed to combine 2D image and 3D point cloud to realize 3D detection of such defects. The grayscale pictures are processed, including the enhancement of the image frequency domain, the regional mark and the regional merge, so as to realize the localization of the defects in the two-dimensional pictures. By using the binocular stereo vision system to scan the remanufactured parts surface, Data, while the two-dimensional defect boundary is converted to three-dimensional defect boundaries, to achieve the point cloud data in the positioning of defects. The experimental results show that the system can effectively identify the defects of shallow spots, and the remanufacturing accuracy is high.