Parking Space Detection Using UAV-Based Detection and YOLO in Smart Cities

Road traffic significantly impacts daily lives and consumes valuable time. This paper focuses on the detection of free parking spaces, particularly in crowded areas like campuses and malls, which typically have designated parking areas. We developed a system which utilizes an unmanned aerial vehicle (UAV) that operates at high altitudes to avoid obstacles. The flight process comprises three steps: (a) takeoff and analysis of target coordinates, (b) navigation between designated spots with brief stops, and (c) return to the starting point for landing. The detection system operates independently of the navigation system and is trained to specify parking spaces and recognize empty or “occupied” ones. It employs deep learning techniques and YOLO-tiny algorithm that is trained on the Darknet framework. While the drone’s flight system was developed in a simulation environment, it is capable of real-world application. Real-world data were used to train and test the parking detection system and the results collected were highly encouraging.