Intelligent data processing technologies for unmanned aerial vehicles navigation and control

Abstract

The paper presents the results of research in the field of the development of technologies for processing heterogeneous information from the unmanned aerial vehicle (UAV) onboard machine vision system with the aim of UAV navigation and control. The main problems of information processing for UAV navigation and control are considered; general tasks to be solved for mission planning and performing are formulated. The key problems of machine vision system are multiband image processing and fusion (both for flight planning and onboard processing), object detection and localization, object tracking, object recognition. Modern methods of object detection, recognition and tracking are analyzed. Advanced techniques and algorithms are compared, and the most effective ones are determined. New original methods are proposed for multiband images fusion based on diffuse morphology. The original methods are developed for deep machine learning, which provide high probabilities of given object detection and recognition. The database of model images for machine learning algorithms is created. The characteristics of the developed algorithms and results of their tests on model and real data are presented.