基于局部拟合平面投影搜索最近点的ICP配准

ICP Registration Algorithm Based on Closet Point Searching in Local Plane Projection

针对传统迭代最近点(iterative closest point,ICP)配准算法迭代次数多、收敛速度慢的问题,研究了基于局部拟合平面投影搜索最近点的改进ICP配准算法,在此基础上采用幂法解算单位4元数,最终在速度和精度上完成了对原始算法的优化。采用C#编程语言将改进的ICP点云配准算法程序化,利用Trimble GX对某大型高层建筑物进行实验,分析了配准过程中4元数参数、旋转矩阵参数和平移参数的变化趋势,并对传统ICP配准算法和改进ICP配准算法的配准残差进行比对,结果表明,改进的ICP配准算法迭代次数少、收敛速度快,研究成果可为现有点云数据配准技术提供参考。

Aiming at tackling the problem that the traditional ICP registration algorithm has many iterations and slow convergence rate, the paper proposes an improved ICP registration algorithm based on local fitting plane projection search to optimize the original algorithm in terms of speed and precision. The improved ICP point cloud registration algorithms programmed in C#, and Trimble GX is used to experiment on a large high-rise building. The variation trends of quaternion parameters, rotation matrix parameters and translation parameters in the registration process are analyzed. The registration residuals of the traditional ICP registration algorithm and the improved ICP registration algorithm are compared. The results show that the improved ICP registration algorithm has fewer iterations and faster convergence. The research results can provide reference for existing point cloud data registration technology.