A real-time crime detection and tracking solution using AI-powered surveillance, built by Team Guardian Eye.
- Abinav S
- Dhilip
- Lokeshwaran
The AI-Powered Smart Surveillance System addresses the inefficiencies of manual CCTV monitoring by automating the process of suspect detection, tracking, and alert generation across city-wide camera networks.
This project was built under the Open Innovation track to enhance real-time crime investigation and improve public safety using computer vision and AI.
Monitoring crime using traditional CCTV systems is manual, time-consuming, and lacks scalability. This project aims to:
- Automate object and person detection.
- Enable cross-camera tracking.
- Provide real-time alerts for suspicious activity.
We designed an AI-driven surveillance system that:
- Integrates multiple CCTV feeds into a unified platform.
- Uses object detection and facial recognition to identify suspects.
- Tracks individuals across different locations using smart tracking.
- Triggers alerts when suspicious behavior is detected.
- YOLO – Real-time object detection.
- DeepSORT – Multi-object tracking.
- dlib + OpenCV – Face detection and recognition.
- NestJS – Scalable backend framework.
- Prisma – Database ORM.
- SQLite – Lightweight relational database.
- React.js – Interactive UI for monitoring and control dashboard.
-
CCTV Integration
Centralized collection of multiple camera feeds into the backend. -
Object Detection
YOLO detects humans and other objects of interest in real-time. -
Cross-Camera Tracking
DeepSORT maintains tracking IDs of individuals across frames and locations. -
Face Recognition
dlib + OpenCV recognize known or flagged faces. -
Alert System
Suspicious movements or entries trigger live alerts for authorities.
- Law enforcement agencies
- Security firms
- Smart city projects
- Rising need for automated surveillance in crime-prone zones.
- Enhances efficiency of investigation with faster response time.
- Integration with city-wide public camera networks.
- Add vehicle tracking using ANPR (Automatic Number Plate Recognition).
- Improve alert intelligence using anomaly detection.
- Deploy on edge devices for low-latency real-time processing.
- ✅ Innovation: Smart surveillance combining detection, tracking, and alerting.
- ✅ Feasibility: Compatible with existing infrastructure.
- ✅ Technical Depth: Combines deep learning, facial recognition, and scalable backend.
- ✅ Social Impact: Assists in crime prevention and enhances public safety.
This project is open-source and licensed under the MIT License.