Three-Dimensional Environment Model Construction from an Omnidirectional Image Sequence

Abstract

When mobile robots execute autonomous tasks, map information is important in path planning and self-localization. In unknown environments, mobile robots must generate their own environmental maps. This paper proposes three-dimensional (3D) environment modeling by a mobile robot. The model is generated from results of 3D measurement and texture information. To measure environmental objects efficiently, the robot uses an image sequence acquired by an omnidirectional camera with wide field of view. The measurement method is based on structure from motion. Triangular meshes are constructed from 3D measurement data. The 3D model is constructed by texture mapping to the triangular mesh, proven by experimental result to be effective.