Generación de mapas 3D globales a partir de primitivas geométricas locales

Abstract

The interest in 3D city models has increased in recent years. Two approaches are predominant in 3D reconstruction: the first is using a stereo system and the second is the feature extraction of the cloud of point provided by the LIDAR. LIDAR technology is emerging as more efficient supply of digital models for environments, since the data is already providing a 3D reconstruction. In this work, the environment has been characterized by the mathematical model. The focus of our research is to model structured environment by planes through the 3D data provided by the LIDAR. The main plane corresponds to the ground and other planes present in the cloud correspond to structured objects, which may be the facades of buildings, sidewalks, walls and more. With different planes and segments of points, we proceed to calculate the clusters belonging to each segment, defined by a plane. The plane equation computing in the modeling of the environment correspond to a set of points. The set of points in the plane can be defined by several clusters. The cluster segmentation is performed using MeanShift, and allows us to subdivide the plane into regions of the plane bounded by line segments. To test the proposed algorithms, we developed a prototype information system using JOGL3, in which various tests were run. The work shows the reconstruction for two acquisitions particular, corner of the plaza Zenea in the historical center of Querétaro and parking CICATA, Querétaro unit.