Visual Navigation Algorithms for Aircraft Fusing Neural Networks in Denial Environments

Area of Science:

• Aerospace Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Aircraft edge computing platforms face limitations in processing power, especially in satellite-denied environments.
• Complex operational scenarios require robust navigation solutions independent of external signals.

Purpose of the Study:

• To develop a lightweight visual navigation algorithm for aircraft edge computing.
• To enhance navigation accuracy and robustness in satellite-denied and complex environments.

Main Methods:

• Fusion of neural networks with visual navigation techniques.
• Utilization of object detection for dynamic object labeling.
• Implementation of dynamic feature point elimination for improved feature extraction.

Main Results:

• The proposed algorithm significantly improves navigation accuracy.
• Demonstrated high robustness compared to existing methods like monocular ORB-SLAM2.
• Validated on both public datasets (TUM) and physical flight experiments.

Conclusions:

• The developed algorithm effectively addresses computational constraints in aircraft edge computing.
• Achieves superior navigation performance and robustness in challenging operational conditions.
• Suitable for real-time applications in satellite-denied scenarios.