The work aims to explore the application of human-machine hybrid intelligence in the aerospace field, with a core focus on enhancing the flight performance and safety. A comprehensive intelligent perception system integrating language and motion perception, physiological modality perception, and cognitive behavior perception was constructed to achieve a full understanding of the pilot's state. By introducing techniques such as feature visualization, local interpretability, and interactive transparency, the transparency and credibility of artificial intelligence decision-making were improved. It provided an in-depth analysis on the characteristics and applicable scenarios of three human-machine collaboration modes:human-in-the-loop, human-on-the-loop, and human-out-of-the-loop. A dynamic human-machine collaborative decision-making paradigm was proposed. This paradigm combined the computational advantages of the machine with the intuitive and creative decision-making capabilities of humans, providing theoretical support and practical pathways for the development of aviation intelligence.