Global Optimization Method On 3D Building Reconstruction From Airborne LiDAR Point Clouds

As the foundation of digital city construction,3D building reconstruction is of a current research interest in surveying, mapping and remote sensing. Because of the rich texture and geometric information, the automatic building reconstruction from images still remains to be a challenging issue. Because of being able to directly collect dense, accurate 3D point clouds of urban objects, airborne LiDAR (Light Detection and Ranging) technology provides an efficient solution to this need. However, most of the reported methods on building reconstruction from LiDAR data are based on local approaches. They tend to fail to reconstruct building models when using noisy and incomplete data. To resolve this problem, this paper presents a global optimization approach to roof segmentation and topology reconstruction from airborne LiDAR data. The details of the approach are as follows:1) A global optimization approach to building roof segmentation. This paper presents a global plane fitting approach for roof segmentation from LiDAR point clouds. Starting with a region growing approach, an initial segmentation is first derived from roof LIDAR points. Such initial segmentation is then optimized by minimizing a global energy function consisting of the distances of LIDAR points to initial planes, spatial smoothness between data points, and the number of planes. As a global solution, the proposed approach can determine multiple roof planes simultaneously. Two LiDAR data sets of YangJiang (Guangdong) and Wuhan university are used in the study. Experimental results show that the completeness and correctness are increased from 77.0% to 92.9% and 92.6% to 97.7,0%, respectively; and the reference cross-lap rate is reduced from 17.7% to 0%. As a result, the incorrect segmentation that often occurs at plane transitions is satisfactorily resolved; and the topological consistency among segmented planes is correctly retained even for complex roof structures.2) A space partitioning of building with roof planes and 3D cell complex. As a discrete optimization method, a volumetric graph needs to be constructed when minimizing energy funcation via Graph cuts. This paper presents a space partitioning of building with roof planes and 3D cell complex. The advantage of this approach is that it can partition building space into volumetric cells with geometric features and topology, from which the volumetric graph for energy minimization via Graph cuts can be derived.3) A global optimization approach to building roof reconstruction. This paper presents a global topology reconstruction approach to building reconstruction from LiDAR point clouds. Starting with segmentation results, initial building geometric features and their topological relations are determined by the intersection of segmented planes. With the help of roof LiDAR points, the 3D cells of scene is labeled as empty or occupied via minimum cut, where the surfaces between empty and occupied cells form the reconstructed model. Two LiDAR data sets of YangJiang (Guangdong) and Wuhan university are used in the study. Experimental results show that the completeness of reconstructed roof planes is 83.7%. Comparing with the existing data-driven approaches, it can improve the topology quality of reconstructed building models. Note that the proposed can create building models with appealing appearance even using incomplete roof point clouds.