Low-cost 3D Building Modeling via Image Processing

In the era of Big Data, 3D electronic maps and GIS are widely used in geographic information analysis and urban computing. However, the production cost of electronic maps is expensive, especially that of 3D modeling of buildings that are important geographic objects. In this paper, we propose a fast and low-cost method for 3D modeling of buildings by using building images provided by online electronic maps. The method employs edge detection to identify the bottom contours of buildings in 2D building images; it then takes DNN-based object detection and semantic segmentation to spill a single building from 3D building images, and calculates the height of the building facades. We use this approach to model 22,740 3D buildings in the urban area of Nanjing, and take a 5m high-resolution electronic map as the benchmark to evaluate the modeling accuracy. The results show that image-based 3D building modeling is able to provide acceptable-precision 3D geometry information of buildings.