As Japan becomes a stock-based society, utilizing existing buildings is increasingly crucial. Although 3D measurements for acquiring building shape data are developing, challenges remain in effectively utilizing point cloud data obtained from these surveys. To this end, it is essential to efficiently extract architectural elements (clustering) from point clouds, which contain large data volumes and numerous unnecessary points, and reduce them to meaningful data. This study proposes a novel method utilizing building posture estimation and demonstrates an automatic clustering technique for floors, ceilings, and walls of building point clouds.