Urban on-street parking monitoring with drone flights

Abstract

Efficient on-street parking management is a critical challenge for modern cities due to its direct impact on traffic congestion, energy consumption, and driver satisfaction. The dispersed nature of on-street parking across large regions makes monitoring difficult using traditional sensors, severely limiting their scalability in dynamic environments. This paper presents a novel, fully automated drone-based framework to extract on-street parking locations and monitor occupancy directly from raw aerial video. Our approach eliminates the need for predefined maps or inaccurate crowdsourced data. Instead, it infers parking masks from observed vehicle trajectories. The framework detects vehicles and tracks their trajectories to isolate stationary parking behavior from active traffic. We then use a spatial-temporal dwell-time scoring technique combined with a clustering algorithm to generate precise parking masks. A real-world case study in Galatsi, Greece, demonstrates the effectiveness of this method, capturing two-day parking occupancy dynamics with high autonomy. Results highlight the potential of drone-based aerial systems to deliver scalable, costeffective, and adaptive solutions for studying urban mobility and monitoring on-street parking without reliance on any prior information.