Reconstructing and Exploring Massive Detailed Cityscapes

Abstract

We present a state-of-the-art system for obtaining and exploring large scale three-dimensional models of urban landscapes. A multimodal approach to reconstruction fuses cadastral information, laser range data, and oblique imagery into building models, which are then refined by applying procedural rules for replacing textures with 3D elements, such as windows and doors, therefore enhancing the model quality and adding semantics to the model. For city scale exploration, these detailed models are uploaded to a web-based service, which automatically constructs an approximate scalable multiresolution representation. This representation can be interactively transmitted and visualized over the net to clients ranging from graphics PCs to web-enabled portable devices. The approach's characteristics and performance are illustrated using real-world city-scale data.