This project integrates ESP32-based IoT hardware, a web application, and machine learning (SVM) to create an intelligent license plate recognition system for automated smart parking. It merges embedded technology, cloud services, and image processing to automate access control in a secure and efficient way.
The ESP32 microcontroller acts as the edge device, handling:
- Sensor Input: RFID, infrared sensors, and camera modules.
- License Plate Detection: Captures vehicle images using camera and sends them to a backend or image processor.
- Firebase Communication: Updates vehicle status to Firebase Realtime Database in real-time.
Built using C#, the WebApp allows:
- Real-time monitoring of parking slots
- Control of servo-based barrier gates
- Display of recognized license plates and timestamps
Used for cloud storage and live synchronization:
- Logs entry/exit data
- Manages slot availability and barrier control status
A machine learning algorithm was developed using MATLAB to process vehicle license plate images based on the following pipeline:
- Input Image:
.jpgimages of license plates. - Pre-processing: Grayscale conversion, binarization, noise reduction.
- License Plate Localization: Detecting character regions.
- Character Segmentation: Isolating individual characters.
- Character Enhancement: Binarization and cropping of characters.
- Character Recognition using SVM:
- Linear kernel SVM
- Trained on 36 classes (A-Z, 0-9)
- Store Output: Recognized plate numbers saved in a
log.txtfile.
Performance (on 26 test images):
- License Plate Detection: 88%
- Character Segmentation: 84%
- Recognition Accuracy: 77%
- Improve image quality and plate recognition under various lighting/weather conditions.
- Integrate deep learning (CNN) for higher accuracy.
- Add mobile app support for user interaction.
- Use edge AI on ESP32-CAM for real-time recognition at the edge.