Deep Learning-Based Classification of Military Aircraft Types from Satellite Imagery

Automatic classification of military aircraft in satellite imagery is a challenging problem with a high risk of error due to the limited pixel area of targets, variations in image resolution and illumination conditions, background complexity, and strong visual similarity among classes. In this study, a deep learning approach is proposed that integrates multi-feature extraction with attention-based focusing in order to strengthen the classification decision. The proposed architecture combines features extracted in parallel from two convolutional backbones with different representation capabilities and enhances the influence of discriminative regions through channel and spatial attention mechanisms. The model is trained using transfer learning to reduce the risk of overfitting caused by the limited number of samples. The method was evaluated on the MTARSI dataset, which contains 20 aircraft classes. Experimental findings show that the proposed approach achieved 99.79% accuracy on the test set and delivered similarly consistent performance in terms of precision, recall, and F1-score. Confusion matrix analysis further reveals that the misclassification rates remained low, particularly among visually similar platform types, and that the attention mechanism contributed meaningfully to inter-class discriminability. These results demonstrate that attention-enhanced multi-backbone representations provide an effective and practical solution for military aircraft recognition applications from satellite imagery.