形貌视觉测量中立体拼接靶标的设计及应用

设计了一种正六棱柱形状的立体拼接靶标,以靶标侧面6个棋盘格的角点作为全局控制点。基于近景摄影测量技术,建立立体靶标的6个单元模型,通过计算模型内摄站间的相对位姿,推导出棋盘格角点在所属单元模型的局部坐标。以公共棋盘格为中介,确立相邻单元模型的坐标系转换关系。建立靶标的全局坐标系于1号棋盘格,推导该棋盘格平面与其像平面间的单应性矩阵,从而确立全局坐标系和1号棋盘格所处单元模型的坐标系的转换关系。依次递推实现全局坐标系和每个单元模型坐标系的转换,进而计算出全部靶标角点的全局坐标,再经光束平差算法获取精确值。以玻璃表面棋盘格的角点间距作为评价指标,拼接精度优于0.15 mm/m。基于立体拼接靶标的拼接试验表明,实体模型表面4个子区域的局部点云可被精确地拼接成整体点云。与基于全局控制点和平面靶标的拼接方法相比,本方法亦具有更高的拼接精度。

Design and application of 3D target for stitching in shape vision measurement

A three dimensional (3D) target for point clouds stitching with regular hexagonal prism shape is designed, and the 6 chessboard corners on the six sides of the target are used as the global control points. Based on close range photogrammetry, six element models of the 3D target are built. Through calculating the relative positions among camera stations in every model, the local coordinates of the chessboard corners in related model are derived. Taking common chessboard as the medium, the transformation relationship of coordinate systems for the neighboring element models is determined. The global coordinate system is built on chessboard number 1, and the homography matrix between the chessboard plane and its image plane is derived, so that the transformation relationship between global coordinate system and the local coordinate system of the element model, in which the chessboard number 1 is located, is determined. Then, the transformation relationship between the global coordinate system and the local coordinate system of each element model is derived one by one; and the global coordinates of all target corners are calculated. Afterwards, the SBA (sparse bundle adjustment) algorithm is used to obtain the accurate values. Taking the distances between the chessboard corners on glass surface as the evaluation index, the stitching precision is better than 0.15 mm/m. The stitching experiment based on the 3D target for stitching shows that the local point clouds of the four sub area on the solid model surfaces can be precisely stitched into a whole point cloud. Compared with the stitching methods based on global control points and plane target, the proposed method has higher stitching precision.