Multicopter Navigation Based on Visual Odometry with Object of Interest Correction Function
摘要
Abstract
Methods of navigation of unmanned aerial vehicles and factors affecting the autonomy and stability of such systems in small unmanned aviation are considered. The method of navigation by visual odometry methods for a single observer is proposed and described, as well as the algorithm of correction of accumulated error by methods of graphical comparison of binarised masks recognized by convolutional neural network and masks of objects in satellite images with regard to their position.