Satellite-based aircraft surveillance has become a very important aspect in the monitoring of air vehicle activities in military operations. The automatic classification of military aircraft from satellite imagery is very challenging. Military operations require an accurate classification of military aircraft from real-time satellite images. Traditional approaches tend to be less accurate because of the complexity of aircraft shapes and time-varied illuminations and occlusions. Hence, this study proposes a new approach to leveraging satellite-based faster R-CNN (Region-Based Convolutional Neural Network) for military aircraft classification. The proposed scheme includes two main phases. In the first phase, preprocessing techniques help improve the quality of satellite images and remove noise, with the result that the aircraft becomes more clearly detected. Using principal component analysis (PCA) for feature enrichment and dimensionality reduction in our methodology, we are paying attention to the improvement of the accuracy of Faster R-CNN to better classify military aircraft in satellite imagery. In the second phase, an optimized Faster R-CNN model with ResNet50 as its backbone for satellite imagery classifies military aircraft accurately. From the study, impressive results were obtained: accuracy 97.07%, precision 97.34%, recall 98.74%, and an F1 score of 98.04%.
Military Aircraft Classification using Resnet50 based Faster R-CNN for Advanced Military Aircraft Surveillance
2025-03-06
742910 byte
Conference paper
Electronic Resource
English
Engineering Index Backfile | 1961
Engineering Index Backfile | 1950
|Aircraft self reports for military air surveillance
IEEE | 1999
|