Back to projects
AI / CV2024

GNSS-Independent Navigation

Visual landmark detection for GPS-denied environments

YOLOv5PythonOpenCVComputer Vision

Overview

A navigation system that uses visual landmark detection to enable navigation in GPS-denied or degraded environments.

Problem Statement

Traditional GPS-based navigation fails in indoor environments, urban canyons, and jamming scenarios. Alternative navigation methods using visual landmarks provide a robust backup.

Technical Approach

Implemented YOLOv5 for real-time landmark detection and recognition. Combined visual detection with known landmark positions to estimate vehicle location without GPS.

Key Features

  • YOLOv5 landmark detection
  • Real-time video processing
  • Landmark-based positioning
  • GPS-denied navigation capability
  • Cross-platform deployment

Challenges & Learnings

Training reliable landmark detectors required extensive data collection and annotation. Achieving real-time performance while maintaining detection accuracy was a key optimization goal.

Outcome

Developed a proof-of-concept system demonstrating visual navigation capabilities as part of the SAYZEK 2024 competition project.