📹 CCTV IntelliGuard

CCTV IntelliGuard is an edge-based smart surveillance enhancement module designed to work with existing CCTV cameras. It detects and classifies human motion in real time while keeping power and processing requirements minimal.

The system uses OpenCV for lightweight motion pre-detection, activating a YOLOv8 model—optimized with quantization and pruning—only when movement is detected. This approach ensures efficient computation without compromising detection accuracy.

It helpes to identify when a person is entering or exiting specific monitored areas such as classrooms, labs, and offices. When an event is detected, timestamped video clips are stored and alerts with clip links are sent to a connected mobile application ( via Firebase Cloud Messaging (FCM) ).

All processing is deployed on a Raspberry Pi, making it a compact, low-power, and cost-effective plug-in solution for upgrading existing CCTV systems.

---

📌 Problem Statement

Most existing CCTV systems continuously record footage without intelligence. This results in:

• No real-time alerts
• Inability to distinguish humans from general motion

---

🎯 Solution

Attach this AI-based external hardware to CCTV camera to:

• Detects human presence and captures a photo/video snippet of the individual.
• Sends the captured videos directly to the device
• Triggers an alarm alert for unauthorized entries.
• Includes a manual alarm off switch for authorized overrides.

---

Project Goals

• Automate monitoring during off-hours.
• Provide visual proof of entry (photo/video).
• Immediate notification to mobile devices and key personnel.
• Easy manual control (alarm ON/OFF button).
• Scalable for deployment across multiple university rooms.

---

🚀 Features

• Real-time detection of human motion
• Works with existing cameras
• Sends video clip / photos with timestamp to a designated device
• Triggers an alarm during restricted hours until manually turned off
• Minimal hardware cost using Raspberry Pi ( May be in future adds Coral USB Accelerator (can accelerate machine learning models)

---

🧰 Hardware Requirements

Hardware	Purpose
Surveillance Camera	For real-time image/video feed
Buzzer / Alarm	To alert security on unauthorized entry
Raspberry Pi	For processing camera feed and sending notifications
Push Button	Manual alarm off override
Smartphone / Tablet	To receive image or video alert

---

🛠️ Software Stack ( Not finalize )

Software / Library	Role
Python / C++ / Node.js	Backend logic and camera interface
OpenCV	Human detection and image processing
Twilio / Telegram API	Sending automated alerts or messages in
	your applications
FFmpeg	For video clipping and compression
MQTT / HTTP Server	Communication between devices
Real-Time Clock Scheduler	To set active monitoring hours

• Frigate: Focused on efficient recording (clips and snapshots) only when motion and objects (like humans, cars) are detected.

---

🏗️ System Idea

---

---

🏗️ System Architecture

---

Real-World Application

• Designed to be deployed in university / office / school premises for surveillance.
• Helps in identifying unauthorized entries or suspicious activities.
• Integrating an automated alarm system tailored for malls and public spaces to reduce the workload and required number of security personnel

---

Future Improvements

• Face recognition for known vs unknown individuals.
• Remote access to live feed via secure portal.

---

Status

-Stage 1: Initial concept & hardware planning
-Stage 2: Camera setup and motion detection
-Stage 3: Notification system development
-Stage 4: System integration and field testing

---

📅 Project Timeline – 14 Weeks

This section outlines the week-by-week development plan for the CCTV IntelliGuard system, focused on real-time human detection, video clipping, alerting, and scheduling.

---

🕐 Week 1: Finalize Requirements

• Define system features and workflow
• Determine active hours, alert rules, and UI expectations
• Confirm software stack and hardware specifications

🧾 Week 2: Procure Hardware

• Purchase Raspberry Pi, relavent cables, SD card, and other essentials
• Ask permission for CCTV

🛠️ Week 3: CCTV feed

• Install necessary libraries: OpenCV, FFmpeg, etc.
• Try to access the cctv feed through the laptop

🤖 Weeks 4–5 – Motion + Human Detection (2 weeks)

• Implement motion detection with OpenCV
• Use motion as a trigger to avoid unnecessary YOLO processing
• If motion detected → run YOLOv8n for object detection.
• Filter results to detect only the person class.
• Draw bounding boxes + labels when a human is detected.
• Draw bounding boxes + labels when a human is detected.

📹👤 Weeks 6–7 – Human Detection with CCTV Feed (2 weeks)

• Stream video from CCTV (RTSP).
• Detect motion, If motion → run YOLOv8s to check for humans.
• Record pre-event + event + after-event video clips with timestamps.
• 08

📟 Week 8 – Hardware & Basic Setup (1 week)

• Set up Raspberry Pi.
• Get HDMI/VNC connection to display Pi’s output on laptop.
• Connect CCTV feed and confirm Pi can do the previous computing

🤖 Weeks 9-10 – Deployment on Raspberry Pi (2 weeks)

• Install OpenCV, Ultralytics YOLOv8, and dependencies on Raspberry Pi.
• Run full pipeline: Motion detection Human detection Record pre-event + event + after-event video clips with timestamps save in rasberrypi
• Optimize for Raspberry Pi

📱 Weeks 11–12 – App (2 weeks)

• Build mobile/desktop app interface for receiving alerts.
• Store clips in Firestore with timestamps.
• Push notifications through the app.

✅ Weeks 13–14 – Final Testing & Documentation (2 weeks)

• Test full pipeline: detection → clip creation → storage → alert.
• Test under real-world conditions (different lighting, multiple people).
• Write final documentation, diagrams, and deployment guide.