Private or Public Parking Type Classifier on the Driver’s Smartphone

Abstract

As the number of vehicles and car owners rapidly increases, how to resolve problems related to on-street parking is gradually becoming one of the most challenging tasks of smart cities. Smart parking solutions exploit existing technology to provide drivers with information about available parking spots, often leveraging physical, expensive road infrastructures. Our goal is to analyze previous user’s parking spots and discern private places from on-street ones. We identified the main characteristics of the private parking spots and trained a machine learning model that clusters them and classifies the driver’s last parking. The applicability of our work is in smart parking systems, which can benefit from this classification in three ways: i) tagging parking spots in the city without the need of sensor infrastructure; ii) ease parking exchange when a public parking spot gets available; iii) predict when a user is approaching a public or a private spot based on his last parking in the same area.