Back to projects
2025

Acoustic UAV Detection System

NVIDIA Jetson NanoTensorFlowResNet / CNNYOLOv5TensorRTPCA9685 Pan-Tilt ControlSignal Processing

Overview

A hybrid drone-detection rig fusing a microphone-array acoustic classifier with a YOLOv5/TensorRT visual tracker across two NVIDIA Jetson Nanos — the acoustic side estimates bearing via direction-of-arrival, while a PID-controlled, PCA9685-driven pan-tilt camera locks onto and follows detected drones in real time.

Key Highlights

  • 92% classification precision at 15–20 FPS
  • Full 360° direction-of-arrival coverage from the microphone array
  • TensorRT-optimized YOLOv5 visual detection at 18–25 FPS (640×640, FP16) driving closed-loop PID camera tracking
  • Basis for a first-authored review of 20+ studies (2018–2024), published at IEEE SCC 2025

Documentation