Deep Learning based Smart Parking for a Metropolitan Area

Abstract

In this study, we have introduced a method for utilizing the maximum parking space available for a metropolitan city. This will result in much lesser traffic congestion due to street-side parking. Furthermore, it will also decrease the hassle drivers face when they have to leave their vehicles on the side of the road to do other activities. The method introduces a Deep Learning based system where parking spaces are detected using Data Capturing Units (DCU). These DCUs feed data into our database which can be accessed by the users from our mobile application. The users can book parking spaces accordingly. All these data are saved in real-time and can be accessed through the mobile application. A vehicle classification system has also been designed that achieves an accuracy of 77% from multiple vehicle classes. Furthermore, a number plate recognition system has been used for the identification and safety protocols of the vehicles in parking sites. The number plate identification system is very precise and achieves an accuracy of over 90% for each digit. To the best of our knowledge, no other system of this kind has been implemented for the city of Dhaka before this. On top of that, successful implementation in a hectic city like Dhaka implies that it can be applied anywhere in the world. We believe this system can have a huge impact in reducing traffic congestions and can save an endless measure of time and money for citizens in a metropolitan area.