Research on fatigue detection of flight trainees based on face EMF feature model combination with PSO-CNN algorithm

Area of Science:

• Aviation safety
• Artificial intelligence in aviation
• Human factors in flight

Background:

• Human factors, particularly pilot fatigue, remain a significant cause of flight accidents despite advancements in aircraft manufacturing.
• Accurate detection of pilot fatigue is crucial for enhancing overall flight safety.

Purpose of the Study:

• To propose and validate a novel model for recognizing pilot fatigue using facial attributions.
• To improve the precision of fatigue detection in flight trainees to enhance aviation safety.

Main Methods:

• Facial attributions were extracted using the Dlib package during flight simulations via the land-air call process.
• A fatigue attribution model (EMF) was constructed using 68 facial landmark points.
• A Particle Swarm Optimization-Convolutional Neural Network (PSO-CNN) algorithm was employed for fatigue recognition training and testing.

Main Results:

• The proposed PSO-CNN algorithm achieved a high recognition accuracy of 93.9% on the test dataset.
• The model effectively pinpointed the fatigue levels of flight trainees.
• The algorithm's reliability was confirmed through comparative analysis with two other machine learning models.

Conclusions:

• The developed fatigue detection model demonstrates high accuracy and reliability in identifying pilot fatigue.
• Implementing this AI-driven approach can significantly contribute to reducing fatigue-related flight accidents.
• This technology offers a promising solution for proactive fatigue management in aviation training.