基于配准残差分布函数点对定权的ICP算法研究

Research on ICP Algorithm Based on Registration of Residual Distribution Function

针对传统ICP配准算法无法抵抗常规粗差点对配准精度的影响问题,研究了基于配准残差分布函数点对定权的改进ICP配准算法,推导了基于配准残差分布函数的点对残差权重值计算公式,在此基础上采用幂法解算单位四元数,最终在速度和精度2个方面完成对原始ICP配准算法的优化。采用C++编程语言将改进的ICP点云配准算法程序化,利用Rigel LMS-Z420i三维激光扫描仪对某雕像进行扫描,通过自编程序对含有常规粗差点的点云数据进行配准实验,将基于点对权重的改进ICP算法与标准ICP算法进行比较,结果表明基于点对权重的改进ICP算法能够有效处理配准数据中存在粗差点的情况,是一种比较精确的抗差配准算法,可对现存配准算法进行有效补充。

Aiming at the problem that the traditional ICP registration algorithm could not resist the influence of conventional rough point on the registration accuracy,we studied the improved ICP registration algorithm based on the registration residual function distribution point-weighting,and derived,the point based on the registration residual distribution function.Based on the calculation formula of the residual weight value,we used the unit method to solve the unit quaternion,and finally completed the optimization of the original algorithm in terms of speed and precision.We programmed the improved.ICP point cloud registration algorithm in C++programming language,and used Rigel LMS-Z420i 3D laser scanner to scan a statue.Then,we performed a registration experiment on the point cloud data containing conventional rough points through selfprogramming,and compared with the standard ICP algorithm.The results show that the improved ICP algorithm based on point cloud weight can effectively deal with the existence of gross error points in the registration data.It is a relatively accurate robust registration algorithm and an effectively complement existing registration algorithms.