Construction of 3-D photorealistic street model by recovering range data and texture data

摘要

We have been investigating a technique for constructing 3-D model of buildings by integrating multi-view range data and texture data. Range data and texture data are incomplete. Range data is acquired with a long range laser scanner (accuracy: ±2.5cm at the distance of 100m), but there are many missing points due to reflection conditions and occlusions. Texture data is taken from various angles, but data is not complete due to occlusions. In this paper, we propose a methodology for constructing 3-D photorealistic street models by recovering range data and texture data by estimating the structure of targets.