Geometric Inference of Building Roof Reconstruction by Fusing LiDAR Data Set and Aerial Photographs

The advent of LiDAR technique has opened up many possibilities for the purpose of object/surface reconstruction. The direct geo-referencing of 3-D laser point towards surface or terrain apparently outperforms photogrammetric method by employing at least two overlapping images in the data acquisition favor. It is, however, still inevitable to see some weaker characteristics of LiDAR system as compared to photogrammetric relatively high ground resolution and richer radiometric content revealed from the images. Therefore, fusion of above two data sets by utilizing mutual strengths for the purpose of reconstructing spatial information can be certainly optimistic and foreseen. The designed steps of building roof reconstruction are carried out by: (1). Hypothesizing 3-D building boundaries from LiDAR data set by edge detection; (2). Constructing the topological relationship of 3-D building boundaries; (3). Back-projecting hypothesized 3-D building boundaries onto the associated photos; (4). Validation of building boundaries by imposing 3-D geometric inference among intersected conjugate lines of candidates found in the images; (5). Merging and adjusting building boundaries. The results show that the proposed system offers more robust and satisfactory building reconstruction result as compared to the situation when only single data set is attempted.