The stringent security measures during aircraft landings pose a substantial challenge for the development of onboard autonomous landing systems, particularly for large aircraft. Vision-inertial fusion techniques are viewed as a pivotal strategy for enabling autonomous navigation during this critical phase. This study aims to investigate and evaluate the potential, navigation capabilities, and accuracy of these methods in the context of large aircraft autonomous landings. To assess the performance, a runway landing simulation environment, runway-simulator, was created, utilizing real flight trajectories from Tibet Nyingchi Airport for realistic visual image simulation. The process involved image detection, sophisticated nonlinear filtering, and an analysis of navigation accuracy. The findings demonstrate that a vision-inertial fusion navigation system utilizing nonlinear filtering can achieve precision on par with Instrument Landing System (ILS) standards during runway landings. However, accuracy varies along the runway direction, with lateral and vertical precision exhibiting high precision in the simulated environment. In conclusion, the proposed vision-inertial fusion method is a key component of an autonomous navigation system for large aircraft during landing, functioning independently of ground support. Nevertheless, its implementation presents several obstacles. Future research and practical applications should delve deeper into the intricacies of visual errors to refine and enhance the effectiveness of onboard landing systems.