颅骨配准是计算机辅助的三维颅面复原技术的重要研究内容之一。颅骨配准的准确与否会直接影响到将来颅面复原的准确性。为此,提出一种新的3D颅骨自动配准算法。该算法考虑到颅骨模型的特殊结构与实现的简便性,首先自动提取颅骨不光滑区域的脊线(Crest lines)以及光滑区域的顶点作为特征点,然后利用迭代最近点(ICP)算法进行粗配准,在此基础上,再采用CPD(Coherent Point Drift)算法对颅骨进行精确配准。实验结果表明,该算法能有效提高颅骨配准的准确性并对缺损颅骨具有一定的鲁棒性。
Skull registration is important in computer-aided three-dimensional craniofacial reconstruction.The accuracy of the skull registration will directly affect the validity of the reconstruction.In the paper,an automatic method for 3D skull registration is proposed.It consists of three steps.First,some points on the crest lines and the smooth surfaces of the skulls are defined as landmarks in consideration of the special structure of skulls.Then,ICP algorithm is applied to roughly align the two skulls.Finally,a fine registration based on the CPD algorithm is implemented.Experimental results demonstrate that the algorithm can effectively improve the accuracy of the skull registration and is robust in the presence of the partial skull.