DroneGuard: Acoustic UAS Detection System

> Note: Above is a capture of the live dashboard identifying a Class 1 UAS threat in a simulated environment.

Project Overview

DroneGuard is a hybrid AI-powered acoustic detection system designed to identify and classify Class 1 Uncrewed Aerial Systems (UAS) in real-time.

Built for the Ottawa Defence Hackathon 2025, this prototype addresses the critical challenge of detecting "Low, Slow, and Small" (LSS) aerial threats that often evade traditional radar cross-sections.

By leveraging Transfer Learning on the YAMNet architecture, DroneGuard acts as a "Shazam for Drones," distinguishing between ambient environmental noise (wind, traffic, birds) and specific drone motor harmonics in cluttered urban environments.

---

Tactical Value: Why This Matters on the Modern Battlefield

As seen in recent conflicts such as the war in Ukraine, the battlefield is increasingly saturated with low-cost, commercial-off-the-shelf (COTS) drones used for reconnaissance and kinetic strikes.

The Challenge

• Radar Blindspots: Class 1 UAS are often too small to generate a reliable radar return, or they fly below the radar horizon in "cluttered" environments like urban centers or heavy foliage.
• Electronic Warfare (EW) Limitations: RF detection relies on the drone communicating with a controller. "Dark" drones flying on pre-programmed paths emit no radio signals, rendering RF detection useless.

The DroneGuard Solution

Acoustic Signature Intelligence (ACINT) serves as the fail-safe layer in a multi-sensor defense grid.

• Passive Sensing: Unlike active radar, this system emits no signals, allowing it to operate without revealing the defender's position.
• Non-Line-of-Sight (NLOS) Detection: Audio waves diffract around buildings and terrain, allowing DroneGuard to hear threats before they are visible.
• Cost Asymmetry: Deploys on standard commercial hardware (microphones/laptops), offering a low-cost countermeasure to low-cost threats.

---

Tech Stack & Architecture

This system utilizes a Microservices Architecture to separate high-performance AI inference from robust mission management.

Component	Technology	Role
The Brain (AI)	Python 3.9+ / TensorFlow	Performs spectral analysis and executes the YAMNet inference engine via a FastAPI wrapper.
The Commander	Java 20 / Spring Boot	Manages sensor state, handles user alerts, and logs threat history using Spring WebFlux.
The Interface	HTML5 / CSS3 / JS	Provides real-time visualization of confidence levels and threat status.

---

Deployment Instructions

Prerequisites

• Java 20 (or JDK 17+)
• Python 3.9+
• Maven
• FFmpeg (Required for processing live microphone audio on Windows)

Step 1: Ignite the AI Engine (Python)

1. Navigate to the AI directory:

   cd drone-ai-python

2. Create and activate the virtual environment:

   # Create (One time)
   python -m venv venv
   
   # Activate (Windows)
   .\venv\Scripts\activate
   
   # Activate (Mac/Linux)
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Start the Neural Network Microservice:

   uvicorn main:app --reload --port 8000

   Status: Listening on Port 8000

Step 2: Launch the Command Center (Java)

1. Open a new terminal and navigate to the backend:

   cd drone-backend-java

2. Run the Spring Boot application:

   mvn spring-boot:run

   Status: Command Center Active on Port 8080

Step 3: Operational Use

1. Open a web browser (Chrome/Edge recommended).
2. Navigate to http://localhost:8080.
3. Click "START MONITORING".
4. Simulate Threat: Play audio of a generic DJI Phantom or Quadcopter into your microphone. The system will trigger a RED alert state upon positive identification.

---

License & Acknowledgments

Developed for the Ottawa Defence Hackathon.

• Model: YAMNet (Google Research).