Study Of Segmentation And Skeleton Extraction Algorithm For Three-dimentional Cerebrovascular Point Cloud Data

Segmentation and reconstruction of point cloud plays an important role in thereverse engineering. An effective segmentation lead to a good reconstruction, then wecan change the reconstructed surface into solid model. The end result can meet theclient’s needs. For the cases of surgical simulation and vascular finite element analysis,people want to get CAD models of these objects. However, for the vessels which havenot been segmented, the resulting model that get from reverse software lacks qualityassurance. Therefore, the paper designed algorithms to complete automaticsegmentation and skeleton extraction and lay a foundation for the subsequent accuratemodeling.First of all, one of the segmentation algorithm is based on principal curvature. Thekd-tree is created for finding of k-nearest neighborhood firstly. Then, calculate thenormal vector of the k-nearest neighborhood. Therefore, the normal curvature of k-nearest neighborhood can be got with the help of normal vector, and the principalcurvature directions can be calculated through the second surface fitting. Finally,according to the similarity of direction of the neighbor’s principal curvature, the initialsegmentation is completed, and then with regional integration method to avoid over-segmentation. In order to solve the problem of the over-segmentation of the curvedvessel and high sensitivity to the raw data, another algorithm that based on connectivityis proposed for the cerebrovascular segmentation. Firstly, r-cube is introduced toconstruct the cover set. Then characterizing cover set with the help of principalcomponent analysis method to determine the starting point of segmentation. At the sametime, designing the criterion of connectivity of cover set to achieve the expansion ofpoints. Therefore, the bifurcation can be determined. With the help of bin, the geometriccenter can be calculated to get the skeleton nodes and to achieve accurate extraction ofthe skeleton. Finally, we can get the segmentation and skeleton extraction resultsthrough Matlab programming. Experimental results show that the algorithm is highaccuracy, robustness and wide application.