改进ICP算法实现多视点云精确配准研究

复杂面形的三维整体测量能否顺利完成取决于不同视下测得的三维点云的配准精度。研究表明:采用点到点,点到三角面配准方法易受噪声干扰,采用面形比较计算量大,且在平面和标准球面情况下容易失效。以粗配准标记点所在的立方体区域为重合区域,使用点到点的多邻接三角面距离最近的点对作为初始匹配点,并根据几何结构最大相似原则对所求得的多个粗匹配点对进行筛选,再对筛选后的点对应用最近点迭代(ICP)算法。改进后的ICP算法实现了重合区域的快速自动定位,实现了不同视下点云的快速精确配准,在多个实例下获得了配准精度优于0.01 mm的实验结果。

Research on precise registration of multi-views based on improved ICP algorithm

The success of 3-D measurement of a complex surface depends on the accuracy of registration of the 3-D data acquired in different views.Study shows the methods of matching by point to point and point to triangular surface can be easily affected by noises,and that by consistency of surfaces may have huge quantity of calculations and fail when the objects are planes or spheres.The iterative closest points(ICP) algorithm is applied on points selected from the overlapping area within the cube where the marking points exist in coarse registration.The point pairs are selected firstly by the closest distance from a point in one view to the adjacent multi-tangent planes of a point in another view,and secondly by the principle of maximum geometric similarity in different views.The improved ICP algorithm makes an automatic and rapid postitioning of the overlapping area,the precise registration of point cloud in different views and its precision is better than 0.01 mm in several experimental instances.