Rapid Modeling of Chinese Huizhou Traditional Vernacular Houses

Abstract

In this paper, we propose an automatic modeling method based on constructive grammar for building Chinese Huizhou traditional vernacular houses, which are famous and important examples of Chinese traditional architecture. The proposed method includes two steps: layout generation and feature generation. In the layout generation step, layout guidelines based on 2-D geometric rules are used to evaluate the layout result. In the feature generation step, a probabilistic network is used to generate the features for a single house while feature conflict penalty functions are used to generate the reasonable features for multiple houses. Both the layout generation and feature generation steps use the Metropolis–Hastings algorithm to search for the optimal solution. To increase the computational efficiency of the proposed system, the two steps are performed on a graphics processing unit device using a Monte Carlo sampler. Moreover, a dynamic probability strategy is used to perform the layout perturbation. The proposed method is easy to implement and extend. The experimental results show that the proposed method is practical and efficient. Non-expert users can rapidly generate reasonable Huizhou vernacular houses via simple and intuitive interactions with the system.