🚨 AI CCTV Intelligence System

An AI-powered surveillance analysis platform that transforms long CCTV footage into a structured, searchable timeline of events. Instead of manually reviewing hours of video, investigators receive AI-generated insights, summaries, and event detection in minutes.

Built using OpenCV, Gemini Vision API, FastAPI, and Streamlit.

📌 The Problem

Modern cities generate massive amounts of CCTV footage every day.

Investigators and security teams often face challenges such as:

• ⏱ Manually watching 6–10 hours of footage
• 🔎 Difficulty identifying critical moments in long recordings
• 🚗 Trouble determining accident causes
• 🕵️ Slow investigation due to lack of structured evidence
• 📂 No easy way to search events inside videos

For example:

After a road accident, police may need to review 8+ hours of CCTV footage just to find the exact moment and cause.

This process is time-consuming, inefficient, and prone to human error.

💡 The Solution

The AI CCTV Intelligence System automatically analyzes CCTV footage and converts it into a timeline of meaningful events using AI.

Instead of watching hours of footage, users can:

• View a timeline of detected events
• Search events using natural language
• Quickly understand what happened and when

It acts like an AI investigator for surveillance footage.

🚓 How It Helps Investigations

This system can significantly assist law enforcement, security teams, and investigators.

Accident Investigation

• Detect vehicles, people, and unusual movements
• Identify the exact moment an accident occurred
• Provide visual evidence timeline

Crime Investigation

• Detect suspicious activity
• Track movement patterns
• Quickly find important moments

Time Efficiency

Instead of watching 8 hours of video, investigators can review:

AI Generated Timeline
---------------------
08:12 AM — White car enters parking area
08:14 AM — Two individuals exit vehicle
08:16 AM — Suspicious movement near gate
08:18 AM — Motorcycle collision detected

This reduces investigation time from hours → minutes.

✨ Key Features

📥 Video Ingestion

Upload CCTV footage in formats such as:

• .mp4
• .mov
• .avi

🎞 Video Frame Processing

Using OpenCV, the system extracts frames efficiently for analysis.

🧠 AI Vision Analysis

Powered by Gemini 2.5 Flash Vision, the system detects:

• 👤 People
• 🚗 Vehicles
• 📦 Objects
• ⚠ Suspicious movements
• 🚨 Possible incidents

🕒 Smart Timeline Generation

AI merges detections and creates a clean chronological event timeline.

Example output:

[
  {
    "time": "00:02:13",
    "event": "Person enters parking lot"
  },
  {
    "time": "00:04:55",
    "event": "Two vehicles collide"
  }
]

🔎 Natural Language Search

Users can search events like:

"Show when the accident happened"
"Find when a car entered"
"Suspicious activity near the gate"

The system instantly finds relevant moments.

📊 Interactive Dashboard

A Streamlit UI dashboard allows users to:

• Upload videos
• Track processing progress
• View summaries
• Filter events
• Search the timeline

⚙️ System Architecture

CCTV Video
     │
     ▼
Frame Extraction (OpenCV)
     │
     ▼
AI Vision Analysis (Gemini)
     │
     ▼
Event Detection
     │
     ▼
Timeline Builder
     │
     ▼
Search Engine + Dashboard

🛠 Tech Stack

Technology	Purpose
Python	Core development
OpenCV	Frame extraction
Gemini Vision API	AI video analysis
FastAPI	Backend API
Streamlit	Interactive frontend
JSON Timeline Engine	Event structuring

📂 Project Structure

AI_CCTV_Intelligence_System/
│
├── backend/
│   ├── api/
│   │   └── api_server.py
│   │
│   ├── ai_analysis/
│   │   └── gemini_analyzer.py
│   │
│   ├── search_engine/
│   │   └── search_engine.py
│   │
│   ├── timeline_engine/
│   │   └── timeline_builder.py
│   │
│   └── video_processing/
│       ├── frame_extractor.py
│       └── video_loader.py
│
├── frontend/
│   └── frontend_app.py
│
├── data/
│   ├── processed/
│   └── uploads/
│
├── .env.example
├── requirements.txt
└── README.md

⚡ Environment Setup

1️⃣ Requirements

Make sure you have:

Python 3.9+
pip
virtualenv (recommended)

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Setup Environment Variables

Create a .env file in the root directory.

GEMINI_API_KEY=your_actual_api_key_here

▶️ Step-by-Step Run Guide

Step 1 — Start Backend Server

uvicorn backend.api.api_server:app --reload

Server will run at:

http://localhost:8000

Step 2 — Start Frontend Dashboard

Open another terminal:

streamlit run frontend/frontend_app.py

Streamlit will open at:

http://localhost:8501

Step 3 — Use the System

1️⃣ Open the Streamlit Dashboard

2️⃣ Upload a CCTV video

3️⃣ Click Upload and Analyze

4️⃣ Wait for AI processing

5️⃣ View:

• 📊 Video summary
• 🕒 Event timeline
• 🔎 Search results

🌍 Real-World Applications

This system can be used in:

• 🚓 Police investigation
• 🏙 Smart city surveillance
• 🏢 Corporate security monitoring
• 🚦 Traffic accident analysis
• 🛑 Crime detection

🚀 Future Improvements

Possible next upgrades:

• Real-time CCTV monitoring
• Vehicle number plate recognition
• Face recognition integration
• Multi-camera tracking
• Crime detection alerts
• AI-generated incident reports

🧠 Inspiration

Security teams shouldn’t waste hours watching silent video footage.

This project aims to transform CCTV systems from passive recorders into intelligent investigative tools.

⭐ If You Like This Project

Give it a ⭐ on GitHub and help make AI-powered surveillance analysis smarter.