Intelligent Data Processing For Navigating Drones

Dmitry Gura, Alexey Khoroshko, Tatyana Sakulyeva, Sergey Krivolapov

This is a review of studies related to onboard data processing with an unmanned aerial vehicle that is aimed at solving a problem of an autonomous drone navigation. The aforementioned kinds of data include but are not limited to machine vision systems, sensors measuring the altitude and speed of the aircraft, and accelerometers. A range of current problems in autonomous drone navigation was established. Original diffuse methods of counterfactual processing of data from the drone’s onboard systems were proposed. A methodology for creating deep learning datasets was developed. The aim of optimizing the procedure of deep learning dataset creation was to achieve high quality object recognition with a constant base (machine vision system). This work showed that the object recognition quality may be optimized drastically by generating a corresponding dataset. The problem of retraining was also considered and measures to reduce this problem were proposed.The study justified the possibilities of creating anonboard drone navigation system based on an object recognition algorithm. An analytical review and systematization of promising object recognition algorithms was conducted with relation to alternative UAV navigation systems.

Volume 12 | 02-Special Issue

Pages: 396-401

DOI: 10.5373/JARDCS/V12SP2/SP20201086