Abstract <p>The paper considers approaches to navigation of unmanned aerial vehicles in the absence of satellite signals using computer vision algorithms. A comparative analysis of absolute navigation methods using the pre-loaded satellite maps is carried out. An algorithm is proposed for the season-invariant visual-inertial navigation based on the image feature matching using the SIFT algorithm with correction based on signals from an inertial measurement unit computationally optimized for the real time use. The results of the algorithm testing using the real unmanned aerial vehicle flight data are presented in comparison with the reference data from a satellite navigation system.</p>