【摘要】 针对异源点云配准存在密度变化、尺度差异、噪声异常和数据缺失等问题,提出了基于高斯混合模型的点云配准模型。通过高斯混合模型描述点云的三维信息,将两片点云的配准问题转化为求解对应高斯混合模型的概率密度估计问题。针对异源点云的噪声,使用能够覆盖点云的最小边界框的体积的均匀分布描述噪声和离群值。选用某水塔作为实验对象,比较基于高斯混合模型的异源点云配准和迭代最近点算法的优劣。实验表明,基于高斯混合模型的异源点云配准在运行时间和精度上都要优于迭代最近点法。更多还原

【Abstract】 Aiming at the problems of density variation,scale difference,large amount of noise and missing data in heterogeneous point cloud registration,a heterogeneous building point cloud registration model based on Gaussian mixture model(Gaussian-based Cross-sourced Registration,GCR) is proposed.The point cloud is described based on the Gaussian mixture model,and the registration problem of two point clouds is transformed into the probability density estimation problem of solving the corresponding Gaussian mixture model.For noise in heterogeneous point clouds,noise and outliers are described using a uniform distribution over the volume of the smallest bounding box of the target point set.A water tower is selected as the test object,and the advantages and disadvantages of heterogeneous point cloud registration and iterative closest point algorithm based on Gaussian mixture model are compared and analyzed.Experiments show that heterogeneous point cloud registration based on Gaussian mixture model outperforms iterative closest point method in terms of runtime and accuracy.更多还原