Integration of image de-blurring in an aerial Mono-SLAM

Abstract

In this article, we discuss the possibility of integrating image de-blurring techniques in an aerial simultaneous localization and mapping by a single camera (monocular simultaneous localization and mapping (Mono-SLAM)). We use an integrated aerial virtual environment together with a six-degree-of-freedom aircraft flight simulator to show the effectiveness of the approach to generate three-dimensional flight trajectories via integration of image de-blurring in the associated loop of the Mono-SLAM. The objective is to increase the overall performance of a flying mission over an unknown area by means of a vision-only method. The integrated aerial virtual environment produces and collects real-time pictures from a nadir-looking vision sensor mounted on the vehicle. Our MATLAB GUI-based toolbox helps user to investigate an offline Mono-SLAM with a predefined de-blurring method integrated with an estimator which extracts navigational parameters. The system efficient architecture allows effective virtual experiments in a completely unknown environment, without using preloaded maps or predefined features. Simulation outcomes demonstrate the feasibility of navigation of aerial robots in inaccessible environments. Different case studies support the conclusion; nonetheless, we observe a number of nonlinearities from de-blurring filters even for a general aviation aircraft.