Computer Engineering and Applications ›› 2008, Vol. 44 ›› Issue (1): 211-215.

• 工程与应用 • Previous Articles     Next Articles

2D-3D pose estimation of industrial sheet metal parts visual inspection

TANG Min1,ZHANG Zu-xun1,ZHANG Jian-qing1,2   

  1. 1.School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,China
    2.State Key Lab for Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,Wuhan 430079,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2008-01-01 Published:2008-01-01
  • Contact: TANG Min

工业钣金件视觉检测中的一种2D-3D自动定位算法

唐 敏1,张祖勋1,2,张剑清1,2   

  1. 1.武汉大学 遥感信息工程学院,武汉 430079
    2.武汉大学 测绘遥感信息工程国家重点实验室,武汉 430079
  • 通讯作者: 唐 敏

Abstract: Present an algorithm including two steps to achieve the automatic 2D-3D pose estimation of sheet metal part.Estimate the initial value by counting the number of edge points along the outskirt of CAD perspective at first;and then to calculate the precision pose by the iteration strategy with the closest straight line based on the generalized point photogrammetry theory.And also present a parallel-optimal 8-neighbour outskirt tracking algorithm.

Key words: 2D-3D pose estimation, generalized point photogrammetry theory, iterative closest straight line, 8-neighbour tracking, parallel algorithm

摘要: 利用图像计算目标物体空间方位一直是计算机视觉领域的重要任务。利用单幅图像结合CAD三维模型实现钣金零件空间方位的2D-3D自动定位。沿CAD三维模型在影像上投影的外轮廓统计影像边缘特征点的数目,根据极值位置确定零件方位的初始参数;提出基于广义点理论的最邻近直线迭代的算法求取零件方位的精确值。为加快轮廓跟踪速度,也对经典的8-领域轮廓跟踪串行算法进行了并行优化。

关键词: 2D-3D定位, 广义点理论, 最邻近直线迭代, 8-邻域跟踪, 并行算法