Airborne Monitoring of Ground Traffic Using Unmanned Aerial Vehicles

Abstract

This article presents an airborne surveillance methodology to monitor the ground traffic and identify suspicious or abnormal behaviours using unmanned aerial vehicles. To track moving targets using an on-board moving target indicator, the tracking filter is designed along with sensor fusion and trajectory smoothing techniques to improve the estimation accuracy. Using the estimated car states, two behaviour recognition algorithms are introduced: (i) string pattern matching to find pre-defined suspicious behaviours in the car driving mode history (expressed as a string of numbers) and (ii) a statistical learning approach with domain knowledge given by road-map information. To verify the feasibility and benefits of the proposed approaches, numerical simulations on moving ground vehicles are performed using realistic car trajectory data from an off-the-shelf traffic simulation software.