Computer vision · Raspberry Pi · Firebase · SQLite
Camera-based parking availability
My final year project. This system detects real-time parking occupancy from camera images and publishes it to a web dashboard for students and SQLite for analysis.
About this project
There are periods of time when finding a parking space in APU Zone B took close to 30 minutes. During these times, I've always asked my friends to take a birds-eye-view image from the top floors of APU to help me find a spot easier. One day I decided to try and automate this.
This system uses an elevated Raspberry Pi Zero 2 W with a Raspberry Pi Camera Module 3 to capture continuous images of the parking lot. It routes the images to my home workstation using SMB over Tailscale, applies homography, detects vehicles, then classifies each slot as vacant or occupied, as shown in the image above.
The detections are then published to Firebase Realtime Database which updates a web dashboard listener, while historical records are stored in a SQLite database for analysis and forecasting.
This project is designed to showcase the full lifecycle of data from hardware-based raw image captures, data preprocessing, to analysis.