An improved extraction method of individual building wall points from mobile mapping system data

Nowadays, since urban planning and landscape simulations are required to get the "characteristics" of target buildings, especially in structures, the reconstruction of 3D building models, necessary for visualization on a computer, becomes crucial. In 3D landscape modeling, a 3D building model that can well represent reality is always the upmost goal. We have already proposed an automatic extraction method of individual building wall points from a Mobile Mapping System (MMS) data. During the process called "plane extraction" to extract multiple planes, a method called RANSAC has been used. However, it is time-consuming to ensure that planes of all walls are extracted. The reason of this problem is the inlier ratio becomes lower as the process goes on. In this paper, we propose a more efficient method which groups the point cloud of a building into "wall groups" based on building structure to increase the inlier ratio. We show that the execution time of our grouping method is about seven times faster than the previous one.