A multi-resolution approach for discovery and 3-D modeling of archaeological sites using satellite imagery and a UAV-borne camera

This paper proposes a method for discovering new archaeological sites from existing satellite imagery, then building a 3-D computer model of those sites using a controlled UAV with an onboard camera. We use an unmanned vehicle and other remote surveillance sensors, coupled with onboard pattern recognition algorithms, to perform a coarse search, and subsequently a fine search to identify structures of interest. We assume the availability of two sensors. The first sensor is a low resolution camera that sweeps an area of interest, such as an imaging satellite. We process the low-resolution image data to identify tentative locations of interest and to provide confidence estimates with this identification. This information is provided to a control algorithm for an unmanned air vehicle, which plans a trajectory to inspect closely promising objects subject to fuel constraints. These close inspections provide sequences of images that are combined to give 3-D reconstructions of the area of interest, leading to accurate classification of the structure. In this paper, we describe the design of the three principal algorithms in this system: machine learning processing of coarse resolution data, the near-optimal path planning subject to fuel constraints, and the high-resolution 3-D modeling from multiple 2-D views of a site. We illustrate the performance of our system on sample LANDSAT satellite data, and using a quadrotor with an on-board camera in a laboratory environment.