DEVELOPMENT OF ALGORITHM TO GENERATE THREE-DIMENSIONAL URBAN MODELS IN THE DENSE URBAN AREAS USING AIRBORNE LiDAR

ABSTRACT: These years in Japan, the demand for three-dimensional (3D) urban model is growing in various applications, such as preparedness to fire spread caused by earthquakes. Airborne LiDAR (Light Detection and Ranging) has been reported to be potential to effectively generate 3D urban model because it measures point clouds containing 3D coordinate which may be easily processed into the 3D model. For this purpose, the technique to classify the point clouds into each house unit is required, but it is found that existing algorithms cannot function effectively especially in the districts where Japanese traditional houses exist because they have almost same height of slant roofs, which are difficult to distinguish from each other. In the present research, an algorithm was developed to automatically generate 3D models using airborne LiDAR, focusing on the accurate extraction of the norms of roof planes and the estimation of roof combination. This algorithm requires tuning of several critical thresholds, i.e. horizontal and vertical distances between houses, and the maximum error to the estimated roof plane. While the system still needs to improve the optimization of such thresholds, it generates 3D models of the Japanese traditional houses more accurately than the existing algorithms.