In response to the issue of pilot fatigue during flight, a system has been developed to identify or predict whether pilots are in a fatigued state while operating aircraft. This innovative system serves as an early warning mechanism for potential risky behaviors that may arise due to pilot fatigue, ultimately aiming to ensure flight safety. The system overcomes the limitations of traditional detection methods by incorporating advanced technologies such as cameras for information acquisition, machine vision, and skin color segmentation algorithms. Specifically, it employs Dlib-based facial landmark detection techniques combined with an improved Lucas-Kanade optical flow tracking algorithm to accurately identify pilot actions and expressions. Additionally, the PERCLOS algorithm is utilized to assess and predict fatigue levels based on specific physiological indicators. By integrating seamlessly with aviation environments, this system effectively reduces the risk of accidents caused by excessive pilot fatigue. It provides a robust solution for enhancing flight safety by offering real-time detection and prediction capabilities, ensuring that pilots receive timely warnings when their fatigue levels increase. This approach not only improves the safety of air travel but also sets a new standard for fatigue monitoring systems in aviation.