Using a Binary Space Partitioning Tree for Reconstructing Polyhedral Building Models from Airborne Lidar Data

During the past several years, point density covering topographic objects with airborne lidar (Light Detection And Ranging) technology has been greatly improved. This achievement provides an improved ability for reconstructing more complicated building roof structures; more specifically, those comprising various model primitives horizontally and/or vertically. However, the technology for automatically reconstructing such a complicated structure is thus far poorly understood and is currently based on employing a limited number of pre-specified building primitives. This paper addresses this limitation by introducing a new technique of modeling 3D building objects using a data-driven approach whereby densely collecting low-level modeling cues from lidar data are used in the modeling process. The core of the proposed method is to globally reconstruct geometric topology between adjacent linear features by adopting a BSP (Binary Space Partitioning) tree. The proposed algorithm consists of four steps: (a) detecting individual buildings from lidar data, (b) clustering laser points by height and planar similarity, (c) extracting rectilinear lines, and (d) planar partitioning and merging for the generation of polyhedral models. This paper demonstrates the efficacy of the algorithm for creating complex models of building rooftops in 3D space from airborne lidar data. Introduction Today, there are increasing demands for rapid and timely compilation of three-dimensional building models from remotely sensed data. Accurate acquisition and frequent up-dating of such models becomes more important source of information for decision making in support of numerous applications, including geospatial database compilation, urban planning, environmental study, and military training (Ameri, 2000). Traditionally, three-dimensional compilation of urban features has been manually conducted under guidance of a human operator using a Digital Photogrammetry Workstation (DPW). Automation of such resource-intensive tasks has been Using a Binary Space Partitioning Tree for Reconstructing Polyhedral Building Models from Airborne Lidar Data Gunho Sohn, Xianfeng Huang, and Vincent Tao a major focus of Photogrammetry and Remote Sensing for many years. In recent years, topographic airborne lidar (Light Detection and Ranging) has been rapidly adopted as an active remote sensing system that uses near-infrared laser pulses (typically 1 to 1.5 mm) to illuminate man-made or natural features on the terrain. The up-to-date lidar system can collect elevation data at a vertical accuracy of 15 cm, at a rate of higher than 100,000 pulses per second. This ability allows the system to produce a dense array of highly accurate and precise three dimensional elevation models, which is a useful property for automating the sophisticated tasks involved in building reconstruction. This paper focuses on the issue of automated construction of 3D building models from lidar data. It is well understood that a general solution to the building reconstruction system entails the collection of modeling cues (e.g., lines, corners, and planes), which represent the major components of building structure. By correctly grouping those cues, geometric topology between adjacent cues, describing a realistic roof shape, can be created. A significant bottleneck hindering the reconstruction process is caused by the fact that extraction of modeling cues is always disturbed by noise inherited from imaging sensors and objects. The most disadvantageous feature of lidar is characterized by irregular data acquisition, which often makes extraction of modeling cues difficult. As shown in Figure 1, the salient boundaries comprising building roof structures, which are easily recognizable in the optical imagery, are often distorted due to a variety of factors, most notably: scanning pattern, point spacing, surface material properties, and object complexity. For this reason, 3D building reconstruction systems have performed most effectively by constraining the knowledge of building geometry either explicitly (model-driven reconstruction) or implicitly (data-driven reconstruction) in order to recover incomplete modeling cues. The model-driven approach pre-specifies particular types of building models so that geometric relations (topology) across modeling cues are provided. By fitting the model to observed data, the model parameters are determined. A good example of model-driven reconstruction was presented by Maas and Vosselman (1999), who were able to determine the parameters of a standard gable PHOTOGRAMMETRIC ENGINEER ING & REMOTE SENS ING Novembe r 2008 1425 Gunho Sohn and Vincent Tao are with the GeoICT Lab, Department of Earth and Space Science and Engineering, York University, 4700 Keele St., Toronto, ON, Canada, M3J 1P3 ([email protected]). Xianfeng Huang is with the State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing (LIESMARS), Wuhan University, 129 Luoyu Road, Wuhan, China. Photogrammetric Engineering & Remote Sensing Vol 74 No. 11, November 2008, pp. 1425–1438. 0099-1112/08/7411–1425/$3.00/0 © 2008 American Society for Photogrammetry and Remote Sensing 07-011.qxd 10/11/08 4:17 AM Page 1425