Abstract

We propose an integrated probabilistic model for multi-modal fusion of aerial imagery, LiDAR data, and (optional)GPS measurements. The model allows for analysis anddense reconstruction (in terms of both geometry and ap-pearance) of large 3D scenes. An advantage of the ap-proach is that it explicitly models uncertainty and allowsfor missing data. As compared with image-based methods,dense reconstructions of complex urban scenes are feasi-ble with fewer observations. Moreover, the proposed modelallows one to estimate absolute scale and orientation, andreason about other aspects of the scene, e.g., detection ofmoving objects. As formulated, the model lends itself tomassively-parallel computing. We exploit this in an efficientinference scheme that utilizes both general purpose anddomain-specific hardware components. We demonstrate re-sults on large-scale reconstruction of urban terrain from Li-DAR and aerial photography data.