论文部分内容阅读
在三维激光扫描技术中,点云数据配准技术直接影响后期建模质量。点云配准主流算法为迭代最近点(ICP)算法,该算法能自动、高精度配准,也具有时间空间复杂度较大、收敛缓慢、易匹配错误对应点等缺点。将基于曲率极值的算法与ICP算法相结合,对曲率特征明显的点云模型进行配准。从算法收敛效率、抗噪性及点云初始位置优劣对算法的影响三方面设计实验,并与经典ICP算法及其他改进算法进行对比。结果表明,该算法对于曲率变化明显的点云数据表现出的收敛效率高于其他算法,对于质量较差的初始数据,该算法收敛稳定性较强。
In 3D laser scanning technology, point cloud data registration technology directly affects the quality of post-modeling. The mainstream algorithm of point cloud registration is an iterative nearest-neighbor (ICP) algorithm, which can automatically and accurately register with shortcomings such as large time-space complexity, slow convergence and easy matching error. The algorithm based on curvature extreme value is combined with ICP algorithm to register the point cloud model with obvious curvature characteristic. The experiment is designed from three aspects: the convergence efficiency of the algorithm, the anti-noise performance and the influence of the initial position of the point cloud on the algorithm, and compared with the classical ICP algorithm and other improved algorithms. The results show that the proposed algorithm has higher convergence efficiency than other algorithms for point cloud data with obvious changes in curvature. For the poor initial data, the algorithm has strong convergence stability.