Vision based navigation of aerial vehicles based on geo-referenced imagery

In this paper, we address a method for terrain aided navigation of aerial vehicles based on vision based sensors which estimate the state of the vehicle (position and attitude) using sequential aerial images. The task is to localize an airborne platform by real-time comparing and matching of an aerial image with a database consists of satellite imageries and digital terrain models (DTM) precisely aligned to geo-coordinates. Such robust matching algorithm is formulated as a rule-based fuzzy image matching algorithm using salient features and descriptors. The corresponding features are then used in a mapping transformation to determine the position and orientation of the aerial vehicle in standard navigation coordinate system. Several experiments were run on aerial image sequences, testifying the robustness and good performance of the implemented matching and pose estimation methods.