Research And Implement Of Representation And Feature Reconstruction Of Complex Geological Surfaces

Surface reconstruction is to give partial information of an unknown surface,and then construct a surface representation that approximates an unknown surface by reconstruction method.Surface representation and reconstruction are widely used in the fields of reverse engineering,medicine,free-form surface design,and film and television entertainment.In the past few decades,the research on reconstruction methods for surface reconstruction in the above fields has been relatively mature.However,for the reconstruction of geologically constrained surfaces,limited by the particularity of geological surface data,there is no major breakthrough in the reconstruction of geological surfaces.First,the data provided during the reconstruction of the geological surface is derived from field survey data,the data is sparse,and the distribution is uneven.Secondly,the geological surface data only has corresponding line data and fault data.These data only have a trend direction for the surface,and there is no complete topology.Furthermore,although the geological surface data is very sparse,in the process of geological survey,it is in the order of kilometers,so the amount of data obtained in the final exploration data is very large.Because of the particularity of geological data,using traditional surface reconstruction methods to reconstruct geological surfaces are not effective.This thesis initialized by the idea of optimizing the seed point search space,and uses traditional methods and neural networks to optimize the interpolation of geological surfaces.The main work and innovations of this thesis are as follows:1.For the traditional reconstruction method,the seed point search is an indistinguishable spatial search problem.In this thesis,the deformed mesh is used for spatial constraint.In the process of geologic surface interpolation,the interpolation space is performed on the orthogonal grid,but the search of the seed points around the interpolation points is done on the deformed mesh.The construction of the deformed mesh is generated by means of a differential iteration through a trend line.The introduction of the deformed mesh makes the attribute of the searched seed point similar to the attribute of the point to be interpolated,and the interpolation result is more accurate than the traditional reconstruction method.2.For the traditional interpolation method,only the spatial correlation problem is used.Based on the point cloud upsampling neural network PU-Net,the joint loss function is used to keep the reconstructed points as possible on the potential surface,and the distribution is kept as uniform as possible.Since the interpolation process uses the feature information of the seed point for interpolation,the interpolation result is more accurate.3.This thesis constructs a software system for the expression and reconstruction of integrated geological surfaces.The software system satisfies the basic interactions required in the process of geological surface expression and reconstruction.The system can complete the surface reconstruction and expression in one system and one process.In this thesis,the actual survey data of a certain work area is used to verify the proposed method,and compared with the reconstruction effect of commercial software.The experimental results show that the two methods proposed in this paper can accurately reconstruct the trend of geological surface,and the reconstruction effect is more traditional.The method works better.