从大规模机载点云中重建几何精确、拓扑正确、语义丰富且屋顶遵循LoD3规范的建筑几何模型是当前机载点云建筑建模的难点和重点。为此，根据建筑几何建模思想，将国内外相关建筑点云建模方法分为5类建模体系，并对每一类体系中的代表文献进行了深入的综述和剖析。在此基础之上，提出当前机载点云建模算法存在的一些共性问题，并给出可能的解决方案及几何建模发展的趋势和后续潜在的研究方向，为完善机载点云建筑重建理论，发展更智能的建模算法，构建更高质量的建筑模型库，提供一定程度的参考和借鉴。

Creating photorealistic building models from large-scale airborne point clouds is an important aspect of urban modeling. Given the complexity of airborne points (i.e., noise, outliers, occlusions, and irregularities) and diversified architectures in the real world, the problems associated with the creation of photorealistic building models pose great challenges, but these problems are not comprehensively addressed by most of the state-of-the-art methods. In this research, intelligent algorithms are developed to create large-scale LoD3 building models with accurate geometry, correct topology, and abundant semantics. The developed algorithms can enhance the abstraction/representation of building point clouds. First, from the perspective of building model mechanism, modeling algorithms are divided into five categories, each of which is reviewed and analyzed in depth. Then, the common problems are determined, and their possible solutions are given accordingly. Finally, the possible directions of future building reconstruction are predicted on the basis of airborne point clouds. We aim to provide beneficial inspiration and relevant references to enhance building modeling theories, develop more intelligent modeling algorithms, and create high-quality building models.