点云数据配准算法的研究

点云配准是点云数据处理的关键，直接影响最后合成结果和模型精度。目前，点云配准方法普遍存在对配准数据初始位姿要求高的缺点。将点云配准分为两个阶段：第一阶段是基于同名点配准，即粗配准，采用人机交互式，配准过程耗时短，节约时间；第二阶段是精配准，在粗配准后，依据最小二乘原理，用间接平差思想，通过最近点迭代算法对点云数据快速配准，并采用目标点集中、目标点坐标与转入目标点集中的点坐标中误差为指标，评价配准精度。进行粗配准的精配准不仅速度快、耗时短，并且可以避免因局部收敛而带来的局部最小问题。试验表明该方法有效可行。

Research on point cloud registration

Registration for point clouds is so important in 3D object modeling that it directly influence the final synthesis result and model precision. The main problem in point cloud registration methods is that the demanding original position registration. The registration for point clouds are divided into two phases. The first phase is coarse registration, by using man-machine interactive and saving time. The second stage is the registration, based on the least-square principle, using the indirect adjustment ideas, through the nearest point iterative algorithm for fine registration and using the target concentration points coordinates into the target concentration between registration accuracy of error as an index in the evaluation. Experimental results show the registration accuracy of this algorithm reaches to a satisfied level. So this experiment method is effective and feasible.