Scene Matching-based Localization of Unmanned Aerial Vehicles in Unstructured Environments

In this paper, we study the feasibility and performance of scene matching-based localization of an unmanned aerial vehicle (UAV) in an unstructured and ambiguous environment. Using a sequence of real time image input and a georeferenced image dataset, we consider a system that combines image similarity evaluation with particle filtering to enable vision- based localization. Through empirical experiments based on aerial imagery of the uninhabited Mitrofania island in Alaska, we present performance and robustness analysis for scene- matching based localization in a perceptually degraded and ambiguous environment.