Automatic Number Plate Recognition and Parking Prediction

A Complete security and control over the Entry with records

Comment

Inspiration

Now days the number of vehicles on the road are increasing day by day. Buildings like shopping complex, Schools, Colleges, etc will have the entry of vehicles all the time. So appointing a security guard is not much efficient. The security guard has to remember the vehicle number of insiders, their entry time, exit time. So I put forward an idea. The Automatic Number Plate Recognition and Parking Prediction System(ANPR and PP sys).

What it does

The ANPR and PP sys is capable of capturing the live visuals using a high quality camera and it processes each frames in real-time and detects the vehicle's number plate. The system will be trained automatically by considering the previous data of vehicle. The parking prediction system is useful for people to know the chances of getting a parking slot. The system is capable of notifying the security guard if any unauthorized vehicles have entered into the campus. Hence the sale of drugs and unauthorized people inside the campus can be prevented. The security guard will get a text message if any outside vehicles are still inside the campus after 15 minutes of allotted time. The guard can see the list of vehicles inside with image of each vehicle.

How we built it

The ANPR and PP sys has front-end and backend. The front end is built using HTML, CSS and JS. The backed is build using Python. Firebase is used for storing the vehicle details like entry time, exit time. owner name, speed, vehicle number, and phone number. The camera will take visuals in real time. each frame is analyzed then to detect a vehicle. If any vehicle is found in a frame, then its number plate is cropped automatically and given to next processing step. In this step the text in the number plate is read using opencv and converted into corresponding text format. Then the system will update a local file with the entry time or exit time, vehicle picture, owner details, etc. Then the system will check the details with the authorized vehicle database. If the vehicle number is not the authorized people database then a 15 minute counter will be triggered. If the vehicle didn't left the campus within allotted time, then a text message with vehicle details will be sent to the guards phone and high volume alarm will be triggered. By learning from the details like count of vehicles inside the campus in each time gap, the system will learn by itself using ML and predicts the parking occupancy for the coming days. These prediction details will be published in a web page which is accessible for public.

Challenges we ran into

I have faced many challenges throughout this project. In order to train the model, I spent almost 3 months for data collection of number plates since the number plate of vehicles comes with different fonts, styles and shapes. It took around 1 month to get the permission from the Motor Vehicle Department to get the API for fetching the vehicle details like owner name, mobile number, etc with vehicle number as input. Bought a high quality camera which was very expensive. The ML training took 56 continuous hours.

Accomplishments that we're proud of

This system has been implemented in my college by me. It has been 4 months the system is still working and protecting my college from unauthorized access and sale of drugs inside the college. Teachers and visitors are using the website to know the parking occupancy and prediction so that they can properly schedule their visiting time.

What we learned

I have learned in ML, Python UI, React, HTML, CSS and OpenCV in depth

What's next for Automatic Number Plate Recognition and Parking Prediction

Next is implementing the system for largescale industries to provide high security to their campus like malls, theater, Companies, etc.

Share this project:

Updates