69. 3D RECONSTRUCTION OF ARCHAEOLOGICAL SITES
Name: Quentin Kevin Gautier
Grad Year: 2019
3D reconstruction of an environment has multiple application from movie making to buildings mapping. It is particularly useful in archaeology to map archaeological sites and artefacts, in order to preserve them digitally. In such cases, we want a high fidelity reconstruction. However, archaeological expeditions to remote areas, such as the Guatemala jungle or an underwater site, are costly and lack access to high-power computing devices. There exist some low-power and low-computing solutions like SLAM (Simultaneous Localization And Mapping), but they lack the precision necessary to accurately reproduce the scanned targets. We are working on applications using low-power depth sensors and SoC like the Google Tango tablet, to improve the quality of such systems. We have built a software capable of reconstructing underground tunnels with accuracy to the order of a centimetre. For more accuracy, Structure-from-Motion is a technique that can reconstruct a 3D static scene from a set of photographs with high precision, and can also be adapted for underwater areas. As it generally requires several hours of computation, it can be difficult to detect on-the-fly if we have enough data to get a good reconstruction. We investigate solutions to analyze the data from a Structure-from-Motion pipeline in order to find missing coverage in the input photographs. By analyzing the reconstructed points at each step of the process, we can focus the algorithm on portions of the scene likely to be missing data, find the "holes" in the 3D structure, and calculate the best poses to take new photos that can fill these holes. This process must be fast enough to run on the field on devices with limited computational power, and eventually can be made real-time to assist users with the data collection.
Industry Application Area(s)
Energy/Clean technology | Software, Analytics