Suitability assessment of ground level imagery for geospatial vr modeling

Improvements in sensor technology and modeling capabilities have resulted in a trend towards generating large-scale VR models of urban scenes, for use in geospatial applications. Selecting the appropriate imagery to generate VR models from a large collection of available ground-level photos is currently performed through visual inspection. In this paper we introduce well-defined metrics to support the automation of this time-consuming task. We make use of geometric and radiometric suitability metrics, describing image suitability for optimal image selection to use in VR modeling. In the paper we present these metrics, and demonstrate the applicability of our approach through experiments.