<p>Aircraft navigation systems must provide a high level of accuracy and integrity, especially during the approach phase, which is one of the most safety–critical stages during a flight. Traditional precision approach guidance systems like Instrument Landing Systems (ILS) and Ground-Based Augmentation Systems (GBAS) rely on costly infrastructure and maintenance, preventing their deployment at under-equipped airports. This paper proposes a Visual-Inertial Navigation System (VINS) for future aircraft precision approach, that integrates a monocular camera with an Inertial Navigation System (INS), by proposing a visual total error model within the current framework of navigation integrity to quantify uncertainties in the vision-based positioning. The system employs a hybrid feature extraction algorithm based on convolutional neural networks to enhance the performance of runway key lines extraction. In correspondence, a visual measurement model is defined, and the potential failures and faults in the visual positioning process are analyzed and defined, leading to the proposal of the visual total error model. The error is conservatively bounded using a folded cumulative distribution function (FCDF) method, enabling dynamic covariance estimation for a generic Kalman filter. Then, a loosely coupled error-state filter fuses visual positioning solution and INS mechanization, achieving continuous positioning required by Category III accuracy (horizontal &lt; 6.2 m, vertical &lt; 2.0 m). Validated through a custom-built Precision Approach Software Simulator (PASS), the system demonstrates achievement of the required navigation accuracy and resilience against initial alignment errors, and the effectiveness of the visual total error model is validated, along with its potential application in future integrity monitoring solutions.</p>