Smart Park

Our Project

Our project aims to solve the problem of unavailability of parking spaces in crowded areas. Live feed is taken from a camera in the parking lot and is fed to a python code which uses opencv to detect any empty slots in parking lots. Then, information about the availability of parking spots is uploaded to a online real time database hosted on firebase. This enables the web application to access this data and inform users about the availability of free parking spots in real time.

The application has an interface that draws the map of virtually any parking area. It is built with JavaScript and canvas. The interface allows us to easily map each parking slot and create a model for visual representation on the web. All slots are custom so it is easy to resize the shapes, delete and name the parking slots in accordance to the different lots around the locality. The map is easily uploaded and stored on Firebase. This map can be then used to view real time availability of the spots in a lot. This map can be embedded in websites and mobile based applications. This being just the first iteration of the map grid tool, in the future we aim to base the map on an AI base for minimal manual intervention.

The website was created using django web frame. The website uses firebase and django's default sqlite for the backend and html, css and javascript on the front end.The home page shoes a picture of the parking lot and a link to see its live feed.

Below we can see the total space in the parking lot the occupied spaces and even the free spaces which is being updated in real time. Below That is a bit about us and what part of the project did we work on. The website uses decorates to block unauthorized or unsubscribed users to access the dashboard or the overview page without a subscription.

The subscribe page handles the payment process where user addition to their card details, enters their email where we send login where they can set up their account for premium use.

The dashboard shows a brief overview of the location. The detail button takes takes the user to a web page with better description of the parking space with details about every parking space and if about their status. This is also updated in real time. The green row indicates free space whereas the red rows indicates occupied space.

Our Aim

We aim to deploy our project to at least one parking slot to test it out. This would enable us to identify other non-foreseen problems, experiment different angles and lighting conditions, and also to calibrate our detection parameters further to increase accuracy. After perfecting the application to one parking slot, we aim to slowly increase our reach. The web application is already set up to accept data for multiple parking lots and display it.

Future Improvements

Currently the car detection system is only optimized for the used video and different lighting conditions might yield a lower accuracy rate. So, we aim to change our approach to detect cars by training a model to detect cars. Furthermore, we have also manually set the coordinates of the parking slots by drawing masks over the picture. This process can also be further improved by using line detection systems to automatically detect and mark parking slots. Our current application implements the use of cryptocurrency. But the user has to manually transfer the amount to the given wallet address. We are about 90% done with integrating the Pi SDK to our application to make payments more efficient.