A Novel Method Based on Topological Perception Theory for 3D Landmark Building Model Retrieval

Abstract

Landmark building models typically exhibit intricate structures, posing challenges for retrieval using generic descriptors. To address this challenge, this paper abstracts and simplifies the model into simple topological structures grounded in topological perceptual theory and introduces a novel categorization of landmark building models. Furthermore, based on view, the quotient distance–angle histogram algorithm is proposed for extracting model topological features. This approach facilitates the identification of intricate landmark building models and enhances retrieval performance. In addition, a new projection method is also proposed. Based on the fact that most landmark building models have correct upright direction and the significant topological features predominantly emerge in the top or side views of these models, this method can accelerate the subsequent extraction and matching of shape descriptors by reducing the extraction of model views. Compared with the existing methods, the proposed projection method is more consistent with human visual perception, and the experimental results verify that the proposed method can improve the retrieval performance when classifying the topological structure of 3D landmark building models.