Procedural modeling of buildings based on façade image segmentation

We propose in this paper a method for the procedural modeling of buildings combing CGA shape grammar with facade image segmentation. The method effectively analyzes the architectural structure and elements and achieves facade image segmentation. Specifically, we generate shape tree of facade image according to image segmentation. We adopt a top-down subdivision scheme to encode rules and a down-top merging scheme to generate 3D models with clear level of detail. Our method can subdivide low-resolution images with certain noise and distinguish prominent architectural elements. We can change the accuracy of models with three-dimensional components so that our models can meet the needs of large-scale urban three-dimensional reconstruction.