Three-dimensional (3D) city modeling has been rapidly increasing the importance due to the need for accurate 3D spatial information for urban planning and management. Building modeling is one of the major tasks in city modeling. Traditionally, building models are mainly reconstructed using aerial images with photogrammetric mapping technique. Because manual measurement of building boundary is required, the photogrammetry approachis time-consuming. Airborne LiDAR data directly provide 3D spatial information and make the reconstruction process of building models to be straight forward and possibly to be very automatic. Focused on the extraction of building boundaries from LiDAR point-cloud data, this paper proposes an algorithm to detect building outlines and to simplify and regulate the outlines to form the structures of building roofs. The algorithmis mainly composed of two parts. The first part is to extract the structure lines of buildings, which can be obtained by calculating the intersections of adjacent planes formed by coplanar point cloud. The second part focuses onthe detection of the building boundaries by using aboundary tracing algorithm. Our preliminary experimental results show the possibility of extracting structure lines and boundaries of buildings. Further study will be continue to refine the algorithm.