针对经典迭代最邻近点(iterative closest point,ICP)算法在三维激光点云配准领域内,存在收敛速度慢、配准误差大、配准效率低的问题,提出了一种基于法向量夹角特征和边界旋转角相融合的改进ICP算法。利用点云区域层划分将点云分成若干独立单元方格,搜寻方格的法向量夹角特征关键点,结合点面曲率对应关系形成初始匹配点对,随后引入距离约束函数,估算边界旋转角和相关动态迭代系数,自动优化刚性变换参数。实验结果表明,与传统ICP算法相比,改进后的算法配准误差降至0.3 %以下,配准时间减少50 %以上,有效提升点云配准效率。
For the classical iterative closest point algorithm in the field of 3D laser point cloud registration,there are problems of slow convergence,large registration error and low registration efficiency. An improved ICP algorithm based on normal vector angle features and boundary rotation angle is proposed. The point cloud is divided into several independent unit squares by using the point cloud region layer division. The key points of the normal angle of the square are searched,and the initial matching point pairs are formed by the corresponding relationship of the point surface curvature. Then the distance constraint function is introduced,estimating the boundary rotation angle and related dynamic iteration coefficients to optimize automatically rigid transformation parameters. The experimental results show that compared with the traditional ICP algorithm,the improved algorithm registration error is reduced to less than 0.3 %,the registration time is reduced by more than 50 %,and the point cloud registration efficiency is effectively improved.