Aircraft Recognition Based on CNN Using Satellite Images

Abstract

This study examines aircraft recognition using Convolutional Neural Networks (CNN) with satellite-derived image data. The research traces the evolution of deep learning, emphasizing the importance of multi-layer neural networks in addressing the limitations of artificial intelligence. The study exclusively utilizes the MTARSI dataset and employs VGG16 and VGG19 models. The motivation stems from the critical role of aircraft recognition in civil aviation, military security, and emergency interventions. The aim of this study is to develop aircraft recognition systems using CNNs. Performance analysis of the VGG16 and VGG19 models in military aircraft recognition tasks demonstrates the superior accuracy of VGG19, with success rates of 82.67% for VGG16 and 89.29% for VGG19. These results highlight the importance of advanced models like VGG19 in the future development of military aircraft recognition systems. The VGG16 and VGG19 models used in this study outperformed other traditional methods. Based on the above analysis, it is evident that the VGG16 and VGG19 models demonstrated higher success rates compared to other traditional methods. The VGG16 model achieved an accuracy of 82.67%, while the VGG19 model achieved an accuracy of 89.29%. These findings underscore the importance of utilizing advanced models like VGG19 in the future development of military aircraft recognition systems, highlighting their significant advantage over traditional approaches in this domain.