Research On Feature Extraction And Registration Of Skull Point Cloud

The computer-assisted craniofacial recovery technique that uses computer graphics to generate the three-dimensional shape of the face based on the geometry of the skull of the unknown face,combined with the geometric relationship between the known skull and the face obtained from the measurement or medical image data,mainly used in archaeology,forensic science and other fields.The registration of the three-dimensional model of the skull is a key step in craniofacial recovery.Due to the complex geometric shape and the high registration of the skull point cloud model,the registration effect of the traditional registration algorithm is insufficient.In this paper,the high-precision registration problem between the skull and the reference skull to be restored during craniofacial recovery is designed.The registration process consists of two steps of initial registration and fine registration is designed to improve the coarse registration and fine registration methods.A cranial point cloud model registration prototype system was implemented to achieve fast and high-accuracy registration of the skull.The main research work of this paper includes:First,an improved 3D skull point cloud registration algorithm based on spin image is proposed.Firstly,the key points of three-dimensional skull point cloud are obtained,and a robust local coordinate reference system is constructed according to the key points and their neighboring points.In view of the shortcomings of single information in spin image,the spin image is calculated and connected separately for each coordinate axis.At the same time,the initial corresponding points are obtained according to the nearest neighbor and k-means algorithm.Finally,the rotation matrix and the translation matrix are solved,which provides better initial conditions for accurate registration.Second,An improved accurate registration algorithm based on ICP is proposed.Aiming at the shortcomings of the traditional ICP algorithm in the initial point sampling and mismatching points,In this paper,the corresponding points are searched according to the number of iterations and the random sampling point set of registration error,which can speeds up the registration speed,In the search process,the distance between the nearest point and the coordinate origin and the angle between the corresponding point normal vectors are added to the corresponding point pair constraint to eliminate the mismatched points.Finally,the algorithm is verified.The experimental results show that the algorithm improves the efficiency of high-precision registration of skull model.Third,A fast skull point cloud registration system is designed and implemented.the system functions mainly include the establishment of skull database,the addition of skull data,the extraction of skull point cloud features,and the registration of skull.Experiments show that the system designed in this paper can find a suitable reference skull for recovery of the skull to be restored.