🦅 FALCON

Automatic Crime Alert and Reporting System

Real-Time Weapon Detection using YOLOv8 • Flask • OpenCV • Raspberry Pi • Twilio

📑 Table of Contents

Overview
Problem Statement
Solution
Features
System Architecture
Detection Methodology
Technology Stack
Hardware Requirements
Software Requirements
Project Structure
Screenshots
Installation
Configuration
Running the Project
Workflow
Model Performance
Security
Limitations
Future Roadmap
Contributing
License
Author

📖 Overview

FALCON (Automatic Crime Alert and Reporting System) is an AI-powered surveillance and emergency response platform designed to detect dangerous objects such as weapons in real-time using computer vision and deep learning techniques.

The system uses a custom-trained YOLOv8 object detection model to analyze live camera feeds, identify potential threats, capture evidence, store detection records, and automatically notify authorized personnel through Twilio SMS alerts.

Unlike traditional CCTV systems that only record incidents and require continuous human monitoring, FALCON performs automated threat detection and rapid response using Artificial Intelligence.

The system combines:

Computer Vision
Deep Learning
Edge AI Processing
Real-Time Video Analysis
Backend Development
Database Management
Automated Alert Generation

The project is designed for real-world surveillance applications including smart security systems, public safety monitoring, and emergency response solutions.

❗ Problem Statement

Traditional CCTV surveillance systems only provide video recording capabilities.

Although cameras continuously monitor environments, identifying dangerous activities still depends on human operators. This creates several problems:

Delayed threat identification
Slow emergency response
Human monitoring limitations
Difficulty reviewing large amounts of recorded footage
Lack of automated incident reporting

In critical situations involving weapons or violent activities, even a few seconds of delay can increase the severity of an incident.

FALCON addresses this challenge by introducing an AI-based automated surveillance system capable of detecting threats, generating alerts, and maintaining digital incident records.

💡 Solution

FALCON provides an automated AI surveillance pipeline:

Camera captures live video input.
OpenCV extracts video frames.
YOLOv8 analyzes each frame for weapon detection.
Detection results are evaluated using confidence thresholds.
When a threat is detected:
- Evidence image is captured
- Detection information is stored
- Incident details are recorded
- Twilio sends SMS alerts
- Dashboard is updated

The system continuously monitors the environment and performs real-time threat detection without requiring constant human supervision.

✨ Features

🤖 AI-Based Weapon Detection

Real-time weapon detection using YOLOv8
Custom-trained object detection model
Bounding box visualization
Confidence score calculation
Image and video evidence capture
Automated threat identification

🎥 Real-Time Surveillance

USB camera support
RTSP/IP camera support
Live video stream processing
Frame-by-frame AI analysis
Raspberry Pi compatible edge deployment

🚨 Automated Alert System

Instant SMS notification using Twilio API
Threat detection timestamp
Detection information sharing
Emergency response notification

📊 Dashboard and Reporting

Web-based monitoring dashboard
Detection history tracking
Incident reports
Captured evidence management
Detection statistics

👤 Authentication System

User registration
Secure login system
Password encryption
Session management
Protected dashboard access

🗄 Database Management

SQLite database integration
Detection record storage
Alert history management
User information management

🖥 Edge AI Deployment

Raspberry Pi compatible architecture
Local AI inference
Reduced response latency
No dependency on cloud processing
Better privacy and security

🏗 System Architecture

                    IP Camera / USB Camera
                            │
                            ▼
                    RTSP Video Stream
                            │
                            ▼
                    OpenCV Frame Capture
                            │
                            ▼
                    Image Preprocessing
                            │
                            ▼
                    YOLOv8 Detection Model
                            │
                            ▼
                 Object + Confidence Analysis
                            │
                            ▼
                    Threat Decision Logic
                            │
             ┌──────────────┴──────────────┐
             ▼                             ▼

     Capture Evidence              Generate Alert
     Image / Video                 Twilio SMS

             │                             │
             ▼                             ▼

        SQLite Database             Emergency Contact

             │
             ▼

       Flask Web Dashboard

             │
             ▼

     Reports and Incident History

The complete system works as a continuous AI surveillance loop:

Camera Input → Frame Processing → Object Detection → Threat Analysis → Evidence Storage → Alert Generation → Reporting

🧠 Detection Methodology

1. Video Input Acquisition

The system receives live video input from:

IP Camera through RTSP protocol
USB Camera

RTSP provides continuous real-time video transmission with low delay.

2. Frame Extraction

OpenCV captures the incoming video stream and converts it into individual frames.

Example:

30 FPS video = 30 individual images processed every second.

3. Image Preprocessing

Before AI processing:

Frames are resized
Pixel values are optimized
Input format is prepared for YOLOv8

4. YOLOv8 Object Detection

The YOLOv8 model analyzes each frame and identifies possible weapons.

The model returns:

Object class
Bounding box coordinates
Confidence score

Example:

Weapon Detected

Confidence: 92%

5. Threat Decision Logic

The system does not generate alerts for every detection.

A confidence threshold is applied:

Example:

Confidence > 60%

AND

Object = Weapon

Then:

Threat Confirmed

6. Evidence Generation

After threat confirmation:

Detection image is captured
Bounding box is stored
Timestamp is recorded
Incident data is saved

7. Alert Generation

The Flask backend triggers Twilio SMS notification containing:

Threat information
Detection time
Camera information

8. Database Storage

Detection information is stored in SQLite:

Date and time
Detected object
Confidence score
Evidence path

🛠 Technology Stack

Layer	Technology
Programming Language	Python
Backend Framework	Flask
Frontend	HTML, CSS, JavaScript
AI Model	YOLOv8 (Ultralytics)
Computer Vision	OpenCV
Deep Learning Framework	PyTorch
Database	SQLite
Camera Streaming	RTSP
Alert System	Twilio API
Dataset Annotation	Roboflow
Hardware Deployment	Raspberry Pi
Version Control	Git & GitHub

🔩 Hardware Requirements

Development System

Recommended:

Processor: Intel i5 / Ryzen 5 or above
RAM: 8GB minimum
Storage: 10GB+ free space
Operating System: Windows/Linux

Edge Deployment System

Raspberry Pi 5

Specifications:

Raspberry Pi 5
8GB RAM
Camera module/IP Camera
MicroSD Card
Power supply
Network connectivity

Camera Requirements

Supported:

USB Camera
IP Camera
RTSP-enabled CCTV Camera

💻 Software Requirements

Python 3.10+
Flask
OpenCV
Ultralytics YOLOv8
PyTorch
SQLite
Twilio API
Git
Visual Studio Code

📂 Project Structure

FALCON/

│ ├── AI/ │ │ │ ├── init.py │ ├── yolo_detect.py │ ├── live_rtsp_detector.py │ ├── train_yolo.py │ │ │ └── models/ │ └── best.pt │ ├── alerts/ │ │ │ ├── init.py │ └── twilio_alert.py │ ├── database/ │ │ │ ├── init.py │ ├── detection_db.py │ ├── alert_db.py │ ├── add_users.db │ ├── detections.db │ └── alerts.db │ ├── media/ │ │ │ └── output/ │ ├── screenshots/ │ ├── static/ │ ├── templates/ │ ├── train_detect_weapons/ │ │ │ ├── data.yaml │ └── dataset/ │ ├── app.py ├── config.py ├── run.py ├── requirements.txt ├── .env ├── .gitignore └── README.md

Screenshots

Login

Register

Dashboard

Detection Report

Alert

⚙ Installation

git clone https://github.com/Astik97/FALCON.git

cd FALCON

Create virtual environment

Windows

python -m venv venv

venv\Scripts\activate

Linux

python3 -m venv venv

source venv/bin/activate

Install dependencies

pip install -r requirements.txt

🔐 Environment Variables

Create a .env file inside the project root directory.

Example:

SECRET_KEY=your_secret_key

MODEL_PATH=AI/models/best.pt

# Twilio Configuration

TWILIO_ACCOUNT_SID=your_account_sid

TWILIO_AUTH_TOKEN=your_auth_token

TWILIO_PHONE_NUMBER=your_twilio_number

ALERT_RECEIVER_NUMBER=receiver_number

▶ Running the Project

python app.py

Open

http://127.0.0.1:5000

FALCON provides a central launcher using run.py.

Start the system:

python run.py

🔄 Workflow


                Start System

                     │

                     ▼

             Camera Initialization

                     │

                     ▼

             OpenCV Video Capture

                     │

                     ▼

              Extract Video Frames

                     │

                     ▼

              YOLOv8 Prediction

                     │

                     ▼

        Weapon Object Detected?

              │              │

             No             Yes

              │              │

              │              ▼

              │       Capture Evidence

              │              │

              │              ▼

              │       Save Image/Video

              │              │

              │              ▼

              │       Store Detection Data

              │              │

              │              ▼

              │       Twilio SMS Alert

              │              │

              └──────────────┘

                     │

                     ▼

            Update Dashboard

                     │

                     ▼

             Generate Reports

📈 Model Performance

Metric	Value
Model	YOLOv8
Dataset	Custom Weapon Dataset
Annotation Tool	Roboflow
Training Framework	Ultralytics
Image Size	640x640
Epochs	50
Precision	85%
Recall	75%
mAP@0.5	81%

Performance may vary depending on:

Hardware capability
Camera quality
Lighting conditions
Detection confidence threshold
Model optimization

⚡ Performance Benchmark

Device	FPS	Latency	Expected Capability
GPU System	15–25	40–60 ms	Faster YOLO inference
Desktop CPU	5–10	120–200 ms	Development and testing
Raspberry Pi	1–3	400–800 ms	Edge AI deployment with optimization

Performance depends on:

YOLO model size
Input resolution
Camera FPS
Optimization method
Hardware acceleration

🔒 Security

FALCON follows security practices to protect system data, credentials, and access.

Authentication Security

User registration system
Password hashing
Secure session management
Protected dashboard access

Data Security

Environment variables used for sensitive credentials
API keys excluded from GitHub
Database access controlled through backend modules
Secure configuration management

AI System Security

Local inference support
Reduced dependency on external cloud services
Evidence stored locally
Controlled access to detection records

Deployment Security

Recommended production improvements:

HTTPS implementation
Firewall configuration
Role-based access control
Database encryption
Secure API authentication

⚠ Limitations

Although FALCON provides automated AI-based surveillance, some limitations exist:

Detection Limitations

Accuracy depends on dataset quality
False positives may occur
Poor lighting can reduce detection performance
Small or hidden objects may be difficult to detect

Hardware Limitations

Raspberry Pi has limited computational power
High-resolution streams require optimization
Multiple camera processing requires additional resources

Environmental Limitations

Camera positioning affects detection accuracy
Network stability is required for RTSP cameras
Extreme weather conditions may affect camera performance

Alert Limitations

Twilio requires internet connectivity
SMS delivery depends on network availability

🗺 Future Roadmap

FALCON is designed as an expandable AI surveillance platform.

Future development will focus on improving intelligence, scalability, and real-world deployment.

Phase 1: System Improvement

Backend and AI Enhancement

Improve YOLOv8 model accuracy
Increase detection speed
Optimize database operations
Improve dashboard analytics
Add advanced reporting system
Improve authentication security

Phase 2: Edge AI Deployment

Raspberry Pi Production Deployment

Deploy complete system on Raspberry Pi 5
Optimize YOLO using ONNX
TensorRT acceleration support
Multi-camera monitoring
Real-time WebSocket communication
Offline edge inference

Phase 3: Advanced AI Surveillance

Intelligent Threat Analysis

Person detection and tracking
Suspicious activity detection
Behaviour analysis
Threat scoring system
Multiple object tracking
Crime pattern analysis

Phase 4: Smart City and Enterprise Deployment

Large Scale Architecture

Cloud deployment
Distributed surveillance network
Mobile application
GPS location tracking
Google Maps integration
Enterprise monitoring dashboard
Smart city security integration

🤝 Contributing

Contributions, issues, and feature requests are welcome.

Steps:

Fork the repository.

Create a feature branch.

Commit your changes.

Open a Pull Request.

git checkout -b feature/new-feature

📄 License

This project is licensed under the MIT License.

You are free to:

Use
Modify
Distribute
Improve

provided that proper credit is given.

👨‍💻 Author

Astik Mohapatra

Backend Developer • AI Systems Developer • Computer Vision Enthusiast

📧 Email:

astikm7007@gmail.com

🔗 LinkedIn

https://linkedin.com/in/astik-mohapatra

🐙 GitHub

https://github.com/Astik97

⭐ Support

If you find this project useful:

⭐ Star the repository

🍴 Fork the repository

📢 Share the project

Built with:

Python • YOLOv8 • OpenCV • Flask • SQLite • Twilio • Raspberry Pi

🦅 FALCON

Automatic Crime Alert and Reporting System

Name		Name	Last commit message	Last commit date
AI		AI
alerts		alerts
database		database
screenshots		screenshots
scripts		scripts
static		static
templates/detection		templates/detection
train_detect_weapons		train_detect_weapons
.env.example		.env.example
.gitignore		.gitignore
README.md		README.md
app.py		app.py
config.py		config.py
requirements.txt		requirements.txt
run.py		run.py

Latest commit

History

Folders and files

Repository files navigation

🦅 FALCON

Automatic Crime Alert and Reporting System

📑 Table of Contents

📖 Overview

❗ Problem Statement

💡 Solution

✨ Features

🤖 AI-Based Weapon Detection

🎥 Real-Time Surveillance

🚨 Automated Alert System

📊 Dashboard and Reporting

👤 Authentication System

🗄 Database Management

🖥 Edge AI Deployment

🏗 System Architecture

🧠 Detection Methodology

1. Video Input Acquisition

2. Frame Extraction

3. Image Preprocessing

4. YOLOv8 Object Detection

5. Threat Decision Logic

6. Evidence Generation

7. Alert Generation

8. Database Storage

🛠 Technology Stack

🔩 Hardware Requirements

Development System

Edge Deployment System

Raspberry Pi 5

Camera Requirements

💻 Software Requirements

📂 Project Structure

Screenshots

Login

Register

Dashboard

Detection Report

Alert

⚙ Installation

🔐 Environment Variables

▶ Running the Project

🔄 Workflow

📈 Model Performance

⚡ Performance Benchmark

🔒 Security

Authentication Security

Data Security

AI System Security

Deployment Security

⚠ Limitations

Detection Limitations

Hardware Limitations

Environmental Limitations

Alert Limitations

🗺 Future Roadmap

Phase 1: System Improvement

Backend and AI Enhancement

Phase 2: Edge AI Deployment

Raspberry Pi Production Deployment

Phase 3: Advanced AI Surveillance

Intelligent Threat Analysis

Phase 4: Smart City and Enterprise Deployment

Large Scale Architecture

🤝 Contributing

📄 License

👨‍💻 Author

Astik Mohapatra

⭐ Support

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages