Skip to content

Vision-based autonomous UAV systems: from visual teleoperation to GNSS-denied global localization

Abstract

The use of Unmanned Aerial Vehicles (UAVs) has significantly evolved in recent years, particularly with advances in onboard computer vision techniques. This dissertation investigates the use of vision as the primary sensing modality to progressively increase the level of UAV autonomy while reducing dependence on external infrastructure and satellite-based positioning systems. Initially, a vision-based teleoperation framework is presented, where the onboard camera serves as the central element for closing the control loop with a human operator. This strategy enables safe task execution even in the absence of direct line-of-sight between the operator and the UAV. The research then advances toward autonomous visual control, integrating segmentation and visual servoing modules to stabilize and align the UAV with visual targets, including compensation for external disturbances such as wind. Subsequently, computer vision is incorporated into higher-level decision-making processes, allowing the identification and selection of suitable landing regions through segmentation and geometric analysis. This stage represents a transition from reactive control to perception-driven autonomy. Finally, the dissertation addresses global localization in GNSS-denied scenarios by proposing an aerial–satellite image matching strategy. The integration of relative visual odometry and satellite-based absolute localization demonstrates the feasibility of achieving global navigation using exclusively visual information. Experimental results confirm that computer vision can support multiple autonomy levels, including assisted teleoperation, closed-loop control, safe region selection, and global pose estimation. The contributions of this dissertation reinforce the potential of onboard vision as a unified sensory framework for intelligent aerial systems, promoting more robust, adaptive, and infrastructure-independent UAV operations.Keywords: unmanned aerial vehicles; computer vision; visual servoing; GNSS-denied localization; visual odometry; aerial autonomy

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.