Facilitating parking management using aerial image recognition

Abstract

Parking problems are an ever-increasing problem, not just in Malta, but around the world. Parking management is a priority since spending a long time searching for a parking space causes frustration for the driver and increases traffic and pollution, damaging the environment around us. Nowadays, technology could be used to ease this problem. This study aims to use a combination of multi-rotors, flight automation, computer vision, machine learning and mobile computing to facilitate parking management. A custom multi-rotor was built, around which an initial study of a flight automation system was created. This multi-rotor was used to collect a dataset of the parking availability over a period of one week. Computer vision techniques were used to automate the process of determining the number of occupied and unoccupied parking spaces. An online survey was carried out with University students, lecturers and administration to collect information on the how they used the parking facilities at the University of Malta, in particular those that were not assessed by the multi-rotor. A Machine Learning model was trained using the dataset captured from the multi-rotor and the survey. Finally, an Android application was developed to represent the percentage of free parking spaces using the Machine Learning model. Statistical analysis of the data collected showed that the computer vision technique used achieved satisfactory results, albeit with some limitations. Several patterns where observed for the parking patterns exhibited in the case study, identifying the ideal time to find a parking space in Car Park 6B at the University of Malta, which the user can visualize through a mobile application.