Method and device for automatic registration of three-dimensional skull

Abstract

The invention provides a method and device for automatic registration of a three-dimensional skull. In the invention, an edge type classifier is trained firstly, then edge extraction is performed on a three-dimensional skull to be registered to obtain three-dimensional hole edges, shape distribution features of all three-dimensional hole edges are extracted, the three-dimensional hole edges are automatically classified into orbit edges, nose frame edges and the like by adopting the edge type classifier, then PCA (Principal Component Analysis) is performed on all edge points on the nose frame edges so as to determine a unified coordinate system, adjustment is performed on the initial position and posture of the three-dimensional skull to be registered, then registration of an edge region is realized according to a corresponding relationship between the three-dimensional skull to be registered and a reference skull edge by adopting a CPD (Coherent Point Drift) algorithm, and finally adjustment is performed on a region with a great registration error of the three-dimensional skull to be registered and the reference skull by adopting a CSRBF (Compact Support Radial Basis Function) so as to achieve accurate registration of the three-dimensional skull to be registered and the entire skull region of the reference skull.

Description

technical field [0001] This application relates to computer technology, in particular to a method and device for automatic registration of three-dimensional skulls. Background technique [0002] Establishing the accurate point-to-point correspondence of the 3D skull model through registration is the premise and basis for the statistical analysis of the skull morphology and the restoration of the skull appearance. However, due to the complex shape and topology of the skull, large individual differences, and the absence of teeth and partial skulls, the automatic registration of 3D skull models has always been a difficult problem and challenge. [0003] At present, the commonly used method for the automatic registration of the 3D skull model is: first use the Iterative Closest Point (ICP) algorithm to perform coarse registration of the 3D skull model, and then perform fine registration on this basis. [0004] However, when the initial positions of the two 3D skull models diffe...

AI Technical Summary

Problems solved by technology

[0004] However, when the initial positions of the two 3D skull models differ greatly, the ICP algorithm is prone to fall into local optimum and cannot be correctly registered.

What's more, the topological structure of the 3D skull model is complex, and there are many holes on the front and side, and the existing ICP algorithm cannot achieve accurate registration of these holes

Images