针对不同视角下的点云配准问题, 提出一种基于改进动态差分进化算法的配准方法。首先利用主成分分析方法估算点云的曲率值与法向量, 并根据法向量计算每个点与其k邻域点的法向量的夹角平均值。然后利用曲率和法向量夹角平均值构造第一特征参数进行第一次特征点提取, 以及利用曲率值构造第二特征参数对点云进行第二次特征点提取。根据得到的特征点云, 最后利用提出的一种基于耦合-最优排序变异的改进动态差分进化算法计算配准参数得到初始配准结果, 再利用改进的迭代最近点算法进行细配准。实验表明, 该配准算法具有配准时间短和配准精度高的优点。

Aiming at the problem of point cloud registration under different angles of view, a registration method based on the Improved Dynamic Differential Evolution(IDDE) algorithm is proposed.Firstly, Principal Component Analysis(PCA) is used to estimate the curvature and normal vector of the point cloud, and the average angle between the normal vectors of each point and its k-nearest neighbors is calculated.Subsequently, the first feature point extraction is conducted by the first feature parameter constructed by the curvature and the average normal vector angle, and the second feature point extraction is conducted by the second feature parameter constructed by curvature.Finally, according to the acquired feature point cloud, the registration parameter is calculated by the IDDE algorithm based on the coupled-optimal ordering mutation proposed in this paper, thus the initial registration result can be obtained, and the fine registration is achieved by an improved iterative closest point algorithm.Experiment shows that the proposed registration algorithm has the advantages of short registration time and high registration accuracy.