A ۳D approach design for aircraft landing on runways using convolutional neural networks

Passenger safety is a top priority in the aviation industry, particularly during the landing phase, which accounts for ۵۳% of aviation accidents between ۲۰۰۵ and ۲۰۲۳. This study introduces a novel approach to aircraft landing by predicting ۳D position components (horizontal, vertical, and distance) through a cost-effective system that eliminates the need for traditional airport equipment. The proposed system integrates image processing techniques with numerical data from navigation systems, aircraft sensors, and airport cameras, leveraging these fused data sources to enhance prediction accuracy. A ResNet۵۰-based Convolutional Neural Network (CNN) is employed, utilizing transfer learning to predict landing parameters with high precision. The novelty lies in the fusion of multimodal data sources and the application of advanced Deep Learning (DL) techniques to predict ۳D position components, which are instantly calculated and displayed on the Horizontal Situation Indicator (HSI). The key preprocessing methods, including data scaling, early stopping, and cross-validation, further enhance the model’s performance by reducing overfitting and improving generalization. The results demonstrate that the proposed model achieves high accuracy (۹۲%), recall (۸۹%), and F۱-score (۹۰%) in predicting landing parameters. These findings contribute to the development of automated landing systems and enhanced passenger safety during landing operations.

کلیدواژه ها:

ResNet۵۰ model ، approach model ، Instrument Landing System ، convolutional neural network