Remote 3D face reconstruction by means of autonomous unmanned aerial vehicles

Abstract

The 3D face model of an individual includes a rich amount of information useful in many application scenarios. Currently, the 3D reconstruction from images is obtained through multiple acquisitions at different distances and angles in a controlled environment. “In the wild’ conditions, or inadequate acquisition systems, can make 3D reconstruction a strongly error-prone problem. This paper illustrates a system that creates 3D face models from images taken by unmanned aerial vehicles in a completely automatic way. The proposed method is adaptive, dynamic, and contactless. No human intervention is required for image acquisition. Experimental results are described—in particular, the comparison between the ideal 3D reconstructions, obtained in controlled and cooperative conditions, and the reconstructions obtained by the drone during flight at different resolutions. The results show good overlap of the models and a low co-registration error. Given the free mobility of the UAVs, the system is suitable, among other applications, for biometrics and open-air access control over large areas.