AI Traffic Violation Monitor System is a real-time computer vision application for detecting red light violations from traffic videos. The system features Automatic License Plate Recognition (ALPR), providing an interactive dashboard that allows users to tune detection parameters, observe results, and view zoomed-in license plate evidence instantly.
- Left panel: Detection and configuration controls (Stop line, AI confidence, Traffic ROI, Road Limits).
- Center: Processed video stream with real-time bounding boxes and violation highlights.
- Right panel: Real-time traffic statistics (Vehicle counts by type) & Live Plate Zoom flash.
- YOLOv8-based detection with persistent tracking.
- Smart Counting Logic: Combines object IDs with position-based fallbacks to ensure accurate vehicle counting even when tracking is temporarily lost.
- Supported vehicle classes: Car, Motorcycle, Bus, Truck.
A violation is recorded when:
- The traffic light state is RED (detected via customizable ROI).
- A detected vehicle crosses the Stop Line while moving in the configured direction.
- Stage 1: Detects the violating vehicle using the primary traffic model.
- Stage 2: Crops the vehicle and passes it to a specialized secondary YOLOv8 model (
yolov8n-plate.pt) to locate the license plate. - High-Res Zoom: Crops the plate area and applies 3x Inter-Cubic Scaling for maximum clarity.
- Real-time Flash: The zoomed plate is flashed on the dashboard immediately upon violation.
- Traffic Direction: Toggle between UP and DOWN flow directions.
- Enforcement Zone (Lane Limits): Define
Lane Left (%),Lane Right (%), andHorizon Limit (Top)to restrict detection to specific lanes and ignore background traffic. - Stop Line Precision: Vertically adjustable stop line for different camera angles.
- Uses an 8-frame buffer to stabilize red/green state detection.
- Reduces false triggers caused by light flickering or temporary obstructions.
- Database Storage: All violations are logged in an SQLite database (
traffic.db). - Evidence Capture: Saves full-frame violation screenshots and cropped plate zooms.
- History Log: Dedicated page to review past violations with time-stamped evidence.
| Parameter | Description |
|---|---|
| Stop Line (%) | Vertical position of the enforcement line. |
| Direction | Direction of vehicle travel (UP or DOWN). |
| AI Confidence | Minimum threshold for YOLO detections. |
| Traffic Light ROI | X/Y/W/H coordinates to monitor the signal lamp. |
| Lane Limits | Horizontal boundaries (Left/Right) for the active road area. |
| Horizon Limit | Vertical boundary; ignore everything "above" this line. |
| Component | Technology |
|---|---|
| AI Engine | Dual YOLOv8 Models (General Traffic + Specialized Plate Detection) |
| Video Processing | OpenCV |
| Backend | Flask + Eventlet (Asynchronous processing) |
| Real-time Comms | Flask-SocketIO (Bidirectional UI updates) |
| Frontend | HTML5, CSS3, JavaScript (Bootstrap 5) |
| Database | SQLite |
| Deployment | Docker & Docker Compose |
git clone https://github.com/hieu-web/traffic-monitor.git
cd traffic-monitor
docker-compose up --buildOpen in browser:
http://localhost:5002
- Install Dependencies:
pip install -r requirements.txt
- Run Application:
python app.py
- Access Dashboard:
Open
http://localhost:5002in your browser.
traffic-monitor/
├── models/ # YOLOv8 weights (best.pt, yolov8n-plate.pt)
├── static/
│ ├── uploads/ # Uploaded video storage
│ ├── evidence/ # Violation & Plate images
│ └── style.css # Dashboard styling
├── templates/
│ ├── index.html # Main dashboard UI
│ └── history.html # Violation log viewer
├── app.py # Flask/SocketIO Server
├── traffic_core.py # Vision processing & Violation logic
├── traffic.db # SQLite database
├── Dockerfile
├── docker-compose.yml
└── requirements.txt