Vector reconstruction, Complex buildings, Orientation recognition and description, Map regularization, Physical geography, GB3-5030, Environmental sciences, GE1-350

Building orientation recognition is a fundamental component of cartographic applications such as map regularization and map generalization. Most current methods effectively calculate orientations for buildings with a single orientation but struggle with complex multi-oriented structures and architectural complexes. This limitation leads to overlooking the spatial relationship between building groups in map generalization. In this paper, we propose a vector reconstruction-based method for calculating building orientation to address the above limitation. By grouping and transforming building edge vectors using a designed vector reconstruction principle, we calculate and correct orientations to obtain the orientation vectors of building footprints. We validate our approach using multiple datasets, and our findings show successful orientation calculation and description for complex buildings and architectural complexes. We also compare our method with the main direction method through map regularization experiments. The results demonstrate that our method better maintains the original shape and geometric structure of buildings, leading to smaller changes in node movement and area. Our approach provides an effective solution for accurately calculating building orientation in complex architectural environments.