A Novel Hybrid Approach Towards Parking Space Occupancy Detection

Abstract

There has been a considerable growth in the volume of road traffic in India. As a result, the demand for parking spaces has increased. Due to less availability of parking spaces, commuters spend many hours on average searching for one. Improper management of the same has lead to overcrowding of roads with vehicles, causing air pollution as well. Thus, to ease the process of locating vacant parking spaces and making this information available for commuters, this paper proposes an end-to-end parking system. The system employs a combined approach, referred to as the Hybrid Model, using Mask R-CNN and YOLOv3 to determine the occupancy status of parking slots and update it in real-time for users to view. A new metric, MDR, was devised to evaluate the car detection performance of Mask R-CNN, YOLOv3, and the Hybrid Model.