Research On Surface Reconstruction Methord For Anomaly Body Of3D Graphics

In the process of conducting geophysical scientific research, geologicalexploration, medical treatment and remote sensing and telemetry, etc, the identification,modeling and surface reconstruction of3D object are often involved. For example, ingeologic interpretation, the identification, tracking and modeling of the objectunderground and horizon structure is very important; another example is that in theprocess of medical care, the3D modeling and surface reconstruction of human organ isalso very important. At present, the main solution to this kind of problem is regardingthe original sampled data as volume data, using volume rendering or the isosurfacetracking method to do3D modeling for the volume data. In this thesis, a new surfacereconstruction method for volume data is proposed, which consider the originalsampled data as point cloud data, so that the surface reconstruction problem of3Dobjects is transformed into reconstruction problem of point cloud surface in computergraphics. Using the theory and methods of computer graphics, the issue has beenanalyzed and studied. Taking3D seismic data as an example, the method in this thesisis simulated, doing the surface modeling for some of the anomaly body in seismic datasuccessfully. In the thesis, the following tasks are completed.Considering the sampling points of geological anomaly body as the data of pointcloud, a new scheme of surface modeling and reconstruction for anomaly body isproposed, which has two following steps.Firstly the density of point cloud data is analyzed and data preprocessing isconducted. By the definition of the concept of the local density, we remove part of theoriginal data which distributed discretely, some noise points are also discarded; thenwe propose an algorithm of space grid dividing, using the relationship between thegrids, we roughly determine which part of sampling points are belong to the surfacearea, finding out the point data shall of the anomaly body.Then the thesis analyzes the surface reconstruction algorithm for scattered data ofboth domestic and foreign papers, comparing their strengths and weaknesses. We findthat the level set theory of implicit surface and the energy function theory is more suitable to solve the problem. We selected the method proposed by Min Wan, Yu Wang,EgilBae et al as the main object of study. Combining with the characteristics of theseismic data, the algorithm has been modified. First of all, we have improved thealgorithm for constructing the data shell of point cloud, replacing tetrahedral shell inthe algorithm with cube shell. This modification makes the algorithm more suitable foranormal body shell data which has lots of noise data; Secondly, We modified thealgorithm of interpolation methods, new interpolation method is more suitable for sandbody sampling data; Thirdly, in the process of using the maximum flow minimum cutalgorithm to solve the energy function, we discard the tetrahedral far away from thesurfaces, because these tetrahedral have little influence for the solving of the problem,removing this part of the tetrahedron, the efficiency of the algorithm is improved.