颅骨点云模型的优化配准

由于颅骨的三维点云数据模型复杂且不同人的颅骨差异较小,对其配准精度要求较高。为了提高颅骨点云模型的配准精度和收敛速度,提出了一种先粗配准再细配准的配准方法。首先,对颅骨点云数据模型进行去噪、简化和归一化等预处理;然后,通过区域划分、区域配准和求解组合系数以及求解刚体变换等步骤实现区域层次上的颅骨粗配准;最后,通过引入动态迭代系数来改进基于旋转角约束的迭代最近点算法,并采用该改进的ICP算法实现颅骨的细配准,从而达到精确配准的目的。实验结果表明:与ICP算法相比,改进的ICP算法的配准精度和收敛速度分别提高了约30%和50%。证明该种先粗配准再细配准的颅骨点云模型配准方法是一种精度高、速度快的有效颅骨配准算法。

Optimization registration of point cloud model of skull

As the three dimensional point cloud model for skull is complex and there is little difference in skulls of different people,it has high requirement for the registration accuracy.In order to improve the registration accuracy and the convergence rate of point cloud model for skull,a kind of registration method of coarse regulation first and fine regulation second was proposed.Firstly,the point cloud model for skull should be subject to de-noising,simplification,normalization and other pretreatments;Then,based on regional partition,regional regulation,solving combination coefficient,solving rigid body transformation and other steps,coarse regulation for skull in regional level was realized;Finally,through introducing dynamic iteration coefficient algorithm,the iterative closest point based on the constraint of rotation angle was promoted;and improved ICP algorithm was used to realize the fine regulation for skull in order to achieve the purpose of accurate registration.The experimental result shows that comparing with ICP algorithm,the registration accuracy and the convergence rate of the improved ICP algorithm are separately improved about 30％ and 50％.therefore,the kind of registration method of point cloud model for skull of coarse regulation first and fine regulation second is an effective skull regulation algorithm with high accuracy and fast speed.