基于平衡VP树的快速空间配准算法

Spatial registration based on balanced VP-Tree

在神经外科导航系统中,空间配准技术是一项关键技术,而确定两点集中点之间的对应关系又是配准中一个不可缺少的环节。为了提高对应点对的查找效率,对经典VP树进行了平衡化处理,并由此提出了一种基于平衡VP树的快速配准新方法。在整个配准过程中,首先用奇异值分解法(Single Value Decomposition,SVD)进行初配准,然后用迭代最近点(Iterative Closest Point,ICP)方法进行精确配准。实验表明:该方法配准快捷、鲁棒性强、配准精度高(配准误差在1mm以内),适用于临床应用。

The spatial registration is a key technology in the image guided neurosurgery and the determination of correspondence position between points of two point sets is also an essential part.A novel image registration method based on balanced VP-Tree （Vantage Point Tree） is proposed,which improves the classical VP-Tree structure＇s balance and enhances the efficiency of searching point-pair.Firstly,SVD （Single Value Decomposition） algorithm is applied for coarse registration.Secondly,ICP （herative Closest Point） algorithm is applied for accurate registration,in which VP-Tree structure is used to find closest point.An experiment shows that the registration is suitable for clinical application,which can be implemented conveniently,quickly,robustly and accurately（errors are limited in 1 mm）.