A Smart Eco-System for Parking Detection Using Deep Learning and Big Data Analytics

Abstract

This work aims to conduct a comprehensive study on existing parking infrastructures and proposes intelligent parking solutions using novel Big Data Analytics with Deep Learning techniques. This research addresses the parking problems faced in most of the cities and growing colleges like University of Alabama in Huntsville (UAH). In this study, segmented images of parking spaces occupied by cars as well as available empty lots were extracted as spatial information and fed to a Convolutional Neural Network (CNN)-based Deep Learning framework. We expect this incorporation of spatial-temporal information would enhance the classification performance of CNNs. For proof-of concept, performance will be validated on a free parking lot dataset-PKLot from Universidade Federal do Parana in Brazil. After desired classification performance is achieved, our robust CNN classification framework can be deployed on a cloud server to classify images in real-time. We also plan to release a parking-support mobile application (Android/IOS) that would display the real-time grid layout of empty and occupied parking spots using GPS information of the user. This app will be constantly updated based on parking information from the cloud-classifier. We expect that the outcome of this study will assist people in managing their time efficiently in finding their closest parking spot. It is also expected to ease the traffic flow and provide better parking management services.