🚦 CampusGuard AI

Automated Traffic Monitoring & Enforcement System

🏆 Top 15 Finalist — IIT Bombay Techfest 2025 (Safetronics: Road Safety Challenge)

---

1. Overview

CampusGuard AI is a computer vision–based system designed to automate traffic monitoring inside university campuses and restricted zones.

Using existing CCTV cameras, the system can:

• Track vehicles in real time

• Estimate vehicle speed

• Detect rule violations such as:

  • Over-speeding
  • Triple riding on two-wheelers

• Automatically read license plates

• Generate violation logs

• Send SMS alerts to vehicle owners

• Provide an administrative dashboard for monitoring

The system eliminates the need for expensive hardware such as radar guns or inductive sensors.

Demo

---

2. Problem Statement

Traffic monitoring inside campuses is usually done manually by security personnel.

This creates several issues:

• Human error
• Limited coverage
• High manpower requirement
• Delayed enforcement
• No structured data collection

Most automated traffic systems require overhead cameras or specialized sensors, which are expensive and difficult to install.

CampusGuard AI solves this problem by working with existing side-view CCTV cameras.

---

3. System Architecture

The project consists of two independent components:

A. Computer Vision Engine (cv_engine)

Processes video feed and detects violations.

B. Django Dashboard (backend_dashboard)

Displays results and manages alerts.

Live CCTV Feed
      │
      ▼
┌─────────────────────────────┐
│   CV Engine (Python)        │
│  YOLOv8 → ByteTrack         │
│        ↓                    │
│  Speed Estimation           │
│  Triple Riding Detection    │
│  Number Plate Recognition   │
└─────────────┬───────────────┘
              │ JSON Log
              ▼
┌─────────────────────────────┐
│  Django Dashboard           │
│  Violation records          │
│  Risk score calculation     │
│  SMS alerts (Twilio)        │
└─────────────────────────────┘

---

4. Core Features

4.1 Speed Detection using Side-View Cameras

Most traffic systems require overhead cameras. CampusGuard AI works with side-view CCTV using a vector cross-product approach, making it deployable on any existing campus infrastructure.

• Two reference lines $L_A$ and $L_B$ are defined in the image plane during a one-time calibration
• Per frame, the system computes the sign of the cross-product of each tracked vehicle's centroid relative to each line
• A sign change indicates a line-crossing event
• Velocity is then computed as:

$$v = \frac{d}{\Delta t}$$

where $d = 4.9\text{m}$ (real-world calibrated distance) and $\Delta t$ is the frame delta at 30 FPS.

---

4.2 Triple Riding Detection

1. Detect all Person and Motorcycle instances per frame via YOLOv8
2. Compute the centroid of each Person bounding box
3. Check strict spatial containment: is the centroid inside a Motorcycle bounding box?
4. If the associated person count exceeds 2 → violation flagged

---

4.3 Automatic Number Plate Recognition (ANPR)

License plates are detected and read automatically.

Pipeline:

Video Frame
     ↓
YOLOv8 Plate Detection
     ↓
Image Crop + Upscale
     ↓
EasyOCR Text Extraction
     ↓
Regex Validation (Indian format)
     ↓
Plate Number Logged

---

5. Technology Stack

Programming Language

• Python 3.10+

Computer Vision

• OpenCV
• YOLOv8 (Ultralytics)
• ByteTrack (Object tracking)
• EasyOCR
• NumPy

Backend

• Django 5.2

Database

• SQLite (development)
• PostgreSQL (scalable deployment)

Notifications

• Twilio API (SMS alerts)

---

6. Implementation Highlights

Feature	Description
Detection model	YOLOv8 (custom trained)
Tracking	ByteTrack (maintains vehicle identity)
Speed estimation accuracy	~5.2% error margin
Camera requirement	Works with side-view CCTV
OCR validation	Indian license plate regex filtering
Architecture	Modular & scalable

---

7. Project Structure

Key modules:

• speed_estimator.py → calculates vehicle speed using geometric crossing logic
• plate_reader.py → extracts license plate text using OCR
• utils.py → risk score calculation and gamification logic
• main.py → runs real-time CV pipeline
• dashboard/ → Django admin interface

---

8. Installation Guide

Requirements

• Python 3.10+
• CUDA-enabled GPU (recommended)

---

Step 1 — Clone Repository

git clone https://github.com/VinitKumarGupta/CampusGuardAI.git
cd campusguardai

---

Step 2 — Install Dependencies

pip install -r requirements.txt

---

Step 3 — Calibration

Define reference lines for speed detection (if using your own test video):

python get_line_coords.py

Update generated coordinates inside:

cv_engine/config.py

---

Step 4 — Run Computer Vision Engine

python cv_engine/main.py

---

Step 5 — Start Dashboard

cd backend_dashboard
python manage.py runserver

---

9. Future Improvements

• Helmet detection
• Night-time optimization
• Cloud database integration
• Multi-camera synchronization
• Real-time analytics dashboard
• Integration with campus security systems

---

_{Built by Vinit Kumar Gupta · PDEU, Gandhinagar}