Research On Volume Data Model Retrieval Method Via Three-dimensional Spatial Feature

With the development of 3D scanning technology and computer graphics technology,the number of 3D volume data models has grown rapidly and been widely used in medical virtual surgery planning,3D geological modeling,mechanical CAD design,meteorology,fluid dynamics simulation and other fields.How to retrieve these volume data models quickly and effectively has an important theoretical significance and practical values.Most of the volume data retrieval methods are based on the geometric and non-geometric features of slices.Note that the essence of these retrieval methods are based on the color,texture and shape of the two-dimensional image slice features,which inevitable result in the loss of the three-dimensional features of volume data,such as spatial shape features and topologies relationship feature,as a result,lots of deviations occurred.To conquer the thorny issue aforementioned,the dissertation focuses on the retrieval method of volume data model.Firstly,the three-dimensional spatial features inside the volume data are constructed by extracting the structural features and shape features of the volume data,and then the similarity of the feature vectors is calculated based on the acquired three-dimensional spatial features to achieve retrieval of the volume data.We achieved the following research outcome:(1)To tackle the barrier that the existing volume data denoising methods cannot effectively remove noise while preserving the boundary sharp feature,we presented an improved WLS volume data denoising method.Modification the regular item to solve the problem of the two-dimensional WLS method cannot effectively remove the boundary noise.Then the improved WLS method was extended to 3D for volume data denoising.An efficient optimization system was used to solve the minimum energy equation for suppressing the boundary noise while preserving the sharp features(2)The existing extraction interface methods can not realize the interface topological relations expression;we proposed an interface topological relationship expression method based on directed skeleton tree.Firstly,the interface point set is obtained by using the gradient magnitude according to the structural features of volume data.We clustered the interface point set with the shape feature and density feature to obtain the sub-interface.Secondly,we constructed a directed skeleton tree from the obtained sub-interface and finished its matrix representation to realize the topological relationship expression of interface(3)As the complex interface is difficult to be fully described,we introduced an interface feature description method based on the skeleton and ridge-valley.Firstly,the cutting method was used to obtain the multi-layer slice of the sub-interface,then each slice is projected and divided into regions calculate the slice center points.The general skeleton is obtained by connecting between the center points.Furthermore,we optimize the general skeleton to obtain the updated skeleton extraction of the sub-interface.In order to obtain the external convex-concave local shape features,the point set of the sub-interface is locally fitted to obtain the principal curvature.By analyzing the obtained the principal curvature,we can calculate the ridges or valleys in each sub-interface.After that,we can accomplish the extraction of the local shape features of the outer convex-concave ridge-valley of the sub-interface.By combining the internal skeleton features and the convex-concave features of external ridge-valley,the description of interface features is overally expressed(4)To solve the problem that the traditional volume data retrieval method has lots of deviation due to the incomprehensive utilization of three-dimensional spatial features,we proposed a volume data retrieval method based on tree structure topology graph.On the one hand,we constructed a tree structure topology graph of volume data and finished its vector representation based on the obtained directed skeleton tree,skeleton and ridge-valley feature On the other hand,the similarity calculation solutions which are from the directional skeleton tree feature,the skeleton feature and the ridge-valley features of the sub-interface are determined.Finally,the similarity of tree structure topology graph vector is calculated to realize the effective retrieval of 3D volume data model.