【作者】 何珍文；

【导师】 王乘；

【作者基本信息】 华中科技大学 ， 空间信息科学与技术， 2008， 博士

【摘要】 地质空间是一个非均质(nonhomogeneous)、非参数化(nonparametic),非直见性(non-eyeable)的三维空间,该空间中的三维地质体动态建模技术正成为三维空间信息技术领域的研究热点问题,也是当今三维空间信息技术中的难点问题之一。本文从三维地质空间的固有属性分析出发,对三维地质空间认知问题、三维地质体数据结构、三维地质建模体系与方法等关键问题进行了分析研究;并在三维地质体数据结构、三维地质体动态构模算法、三维地质体模型的插值拟合算法、三维地质体模型的简化算法以及三维混合空间索引方法等方面进行了深入探讨,提出了多种有效的算法,本文的主要研究内容包括:(1)通过对地质空间性质的归纳总结,提出地质空间的“非均质性、非参数化,非直见性”的三个特性和三维地质空间的“混合空间认知模型”;并基于该空间认知提出了“EBRIM集成数据结构模型”。(2)对三维地质空间中基于钻孔、基于剖面和基于散点的三维地质体动态建模方法进行研究;提出了“基于钻孔的连续地层序列匹配动态建模算法”,该算法通过首先对整个研究区地层出现情况进行分析判断,自动生成该研究区域的标准连续的地层字典,然后通过钻孔与字典的按照一定规则的动态匹配,实现钻孔地层自动对比与连接,有效的解决了在有断层和地层尖灭情况下基于钻孔的自动建模问题。(3)提出了“基于非共面剖面拓扑推理的三维地质体动态重构算法”;该算法将拓扑推理引入三维地质体动态建模过程中,实现了无拓扑变化、地层尖灭、地层分叉情况和断层滑移等四种情况的拓扑推理自动判别与剖面自动对比构模,实现了这四种情况下基于任意剖面序列的三维地质体自动重构。(4)提出了“基于凸包剪切与限定散点集剖分的动态重构算法”;该算法首先将地质钻孔或剖面等数据进行离散插值,并将相关地质信息作为边界限定条件附加在散点集上,然后对离散点集合进行地质年代分类,并计算每类点集的最小凸包,再对这些凸包集按照地质年代顺序相互循环裁剪并进行剖分得到混合地质体模型,最后建立模型拓扑关系并进行模型面片与体元简化。该算法在实现地质体自动构模的情况解决了面模型不支持自动构模,而支持自动构模的体元剖分构模又由于数据量巨大而不具有实用性的问题。(5)提出了“基于TIN和NURBS的虚拟钻孔插值算法”,该算法将NURBS曲面反算插值引入虚拟钻孔插值中,改进了以往单纯的基于TIN的虚拟钻孔插值算法构面比较粗糙的问题;提出了基于剖面拓扑推理的虚拟剖面插值算法,较好的解决了在无拓扑变化、地层尖灭、地层分叉情况和断层滑移等四种情况下虚拟剖面自动生成问题。(6)提出了任意维通用网格GM以及“基于Simplex-Collapse的GM简化算法,GMS”,实现了多种不同网格的统一简化算法,同时基于Simplex-Collapse的GMS比以往基于Edge-Collapse的简化算法具有更高的简化效率。(7)通过对R-Tree,Packed R-Tree,R+-Tree,R*-Tree以及Hilbert R-Tree的分析比较,提出了一种新的索引方法,CSR-Tree(Clustered Sorting Record Tree)索引方法;该算法首先通过空间对象距离相似性聚类,然后对聚类的各个分量集合进行X、Y、Z方向的扫描排序,选取两两距离累积最小的方向对当前分量进行升序排序;然后在对分量集合的中心点进行X、Y、Z方向的扫描排序,选取两两距离累积最小的方向对分量集合的集合进行升序排序,最后递归构造CSR-Tree。该算法充分考虑了空间对象的相邻相关性,有限减小了节点矩形面积,降低了节点矩形交叠概率,提高了R-Tree的查询效率。(8)提出了Grid+CSR-Tree三维混合索引方法(Mix Grid Clustering Sorting Record Tree, MGCSR-Tree);该方法通过两级索引机制将大量空间对象的索引项有机地组织到各个桶文件及其对应的MGCSR-Tree中,既降低了存储开销,又提升了索引的操作效率。上述算法与方法已多个实际项目中使用,取得了较好的实际效果;证明了上述算法的有效性和实用性。更多还原

【Abstract】 Geological space is a three-dimensional space which is nonhomogeneous, nonparametic and non-eyeable. The dynamic modeling of the 3D geological solid in this space is becoming one of the focus problems of 3D spatial information technology research fields, and is also one of the difficult problems in these fields. With the analysis of the geological space attributes, some key problems such as the 3D geospatial cognition, 3D geological solid data structure, the system and methods of 3D geological modeling, the interpolation and fitting algorithms of 3D geological model, the 3D geological solid simplification algorithms and the spatial index, are researched in the paper. The main study contents composed of following items:(1)With concluding and summarizing the geological space attributes, the three basic attributes--nonhomogeneous, nonparametic and non-eyeable, and the mix spatial cognition model of the 3D geological space are presented. Based on this cognition, the extended boundary representation integration model (EBRIM)is presented in this paper.(2)The dynamic modeling methods of 3D geological solid based on borehole, section and disperse points, are studies on. The dynamic modeling algorithm based on continuous borehole stratum series matching is presented. This algorithm includes two main steps. The first is generating the continuous standard stratum dictionary automatically by analyzing and judging the stratums in the research area. The second is doing dynamic matching between borehole and dictionary with guiding by some certain rules. It resolves the dynamic modeling problems in the condition of having fault and stratum annihilating.(3)The dynamic modeling algorithm of 3D geological solid based on noncoplanar section topology reasoning is presented. This algorithm imports the topology reasoning into 3D geological dynamic modeling, realizes the topology reasoning, judging and section corresponding automatically for the four cases: having no topology change, stratum annihilating, stratum bifurcating and fault slippage. Furthermore the automatic reconstruction of 3D geological solid for the 4 cases is implemented.(4)The dynamic reconstruction algorithm based on convex hull cutting and restrictive disperse points division has been presented. This algorithm does interpolation of borehole and section data attached with geological information as boundary restrictive conditions at first. Then the disperse points are classified with the geological age, and the minimums convex hull of each classified points are calculated. Then the convex hulls are sorted by geological age and cutted the adjacent convex hull in sequence, and then the results are divided to receive the 3D geological solid based on EBRI model. The topology relationships are built and the model simplifications are done at last. It resolves the problem that the surface model can not support auto-reconstruction and the solid model can do it, but is not practical because of the huge data.(5)The virtual borehole interpolation algorithm based on TIN and NURBS is presented. It imports the NURBS surface reverse interpolation into this algorithm in order to solve the rough surface problem brought on the virtual borehole interpolation algorithm which only bases on TIN. The virtual section interpolation based on section topology reasoning is presented. This algorithm implements the virtual section auto-generation for the four cases: having no topology change, stratum annihilating, and stratum bifurcating and fault slippage.(6)The n-dimensional general mesh (GM)and the GM simplification (GMS)based on the simplex-collapse are brought out, and implement many kinds of grid general simplification algorithm. The GMS is more effective than the simplification algorithm based on edge-collapse.(7)A new index algorithm named clustering sorting record tree (CSR-Tree),which is not same as R-Tree, packed R-Tree, R+-Tree, R*-Tree, is presented and implemented in the paper. This algorithm includes three main steps. The first step is spatial objects center distance clustering. The second step is sorting the clustering results by scanning X, Y, Z axis, and selecting direction which has the minus sum of the adjacent distance between each two spatial objects. The last step is constructing CSR-Tree recursively. This algorithm reduces the probability of the covering of node boundary rectangle by thinking over the spatial objects’distance relativity, and improves the R-Tree searching efficiency.(8)A Grid+CSR-Tree 3D mix index method named mix grid clustering sorting record tree (MGCSR-Tree)is brought out. This algorithm uses two level index mechanisms to manage huge spatial data, which reduces the memory spending and enhances the index operation efficiency.The algorithms above all has been used in many projects, and obtained good practice effects, and proved these algorithms are valid and practicable.更多还原