Research On 3D Modeling Of Building Based On Multi-Data Source

The concept of the Digital Earth was first put forward in the report The Digital Earth: A21 st century way to understand our planet by Gore, the vice president of US, in 1998. In the last 20 years, both theory and technology have made great progress. In the process of human social development, buildings carry lots of information. It will be better to plan and manage our city by analyzing, concluding and integrating spatial information. The three-dimensional(3D) modeling of buildings become the carrier in the process. Buildings of both regular-shaped and irregular-shaped constitute cities, of which regular-shaped ones are in the majority. Thus, this paper compares and analyzes the proposed methods of 3D modeling of regular-shaped buildings in the development of smart cities, and offers reasonable advices for3 D modeling of massive regular-shaped buildings.This paper studies on massive regular-shaped buildings. Data used for modeling 3D buildings in this research includes surveying and mapping data, images, airborne point clouds and some other data. Following three methods were mainly studied in this research:(i) using surveying and mapping data and buildings’ texture, depending on the actual structure of buildings, making 3D modeling one by one,(ii) using highly overlap building images,according to the structure from motion algorithm to make models,(iii) using airborne point clouds, according to improved Axelsson iterative triangulation algorithm to filter, then vectorization and texture mapping, and the models complete. Comparisons of the three methods mentioned above were conducted on modeling efficiency, automation, visual verisimilitude, operating difficulty and applicability. Different performances were revealed and discussed. Method(i) and method(ii) have advantage in visual verisimilitude. And method(iii) has advantage in efficiency obviously, while has no big difference with other two methods in other aspects.Following advices could be drawn from the experiments and comparisons in this research for 3D modeling of massive regular-shaped buildings in developing smart cities.Airborne point clouds should be done following method(iii) is more available for widelyrange of regular-shaped buildings, while depending on the actual structure or using structure from motion algorithm works better for more complex buildings.