UAV navigation based on videosequences captured by the onboard video camera

Abstract

We propose an approach to navigation for an unmanned aerial vehicle based on finding elements of motion (EM) (linear and angular velocities) by processing the field of local velocities for the motion of an image taken by an onboard video camera. The field of velocities for the image motion, the so-called optical flow (OF), is a linear function of the EM, which allows to use it to find the latter and thus provides an additional way of navigation, which can be rather efficiently used for certain specific problems solved by an UAV in autonomous flight. In this work, we use an algorithm for computing the OF for a given motion of the vehicle (direct problem) and show how to reconstruct the motion of the vehicle by OF observations with methods of statistical estimation (inverse problem).