3D reconstruction of architecture appearance: A survey

Architecture is one of the main elements in outdoor scene of real world, and the reconstructed buildings models have wide applications in virtual reality, computer animation, urban planning and etc. In this paper, we first summarize and classify the recent methods according to different acquisition and modeling principle, discuss and compare the rule-based, image-based and point cloud based reconstruction method respectively from automation degree, input and output data. The paper evaluates the characteristics of each method including advantages and disadvantages and point out that point-based building reconstruction method will be the future research trend. Furthermore, it analyzes the issues such as data loss, mass data processing, manual interaction and visual comprehension that existed in point cloud based method, and the corresponding solutions are proposed so as to improve the integrity, real-time and practicability of the building reconstruction, and also it emphasizes the feasibility by combining the building rules and high-level semantic to reconstruct the details of building model. Finally, the challenge and future research direction are underlined based on the analysis of new progress in this field. © 2013 by Binary Information Press.