Related Experiment Video
Updated: Jun 14, 2026
08:16
Collecting and Processing Drone-based Remotely Sensed Data for Use in Forest Recovery Monitoring
Published on: October 24, 2025
Deep feature extraction and explanation consistency analysis for fine-grained aircraft type classification in remote
Jiachen Liu1, Lei Dong2, Zijing Sun3
1School of Equipment Management and Unmanned Aerial Vehicle Engineering, Air Force Engineering University, Xi'an, 710051, China.
Scientific Reports
|June 12, 2026
Summary
This study introduces an explainable aircraft classification framework (EC-AT) using transfer learning for improved remote sensing analysis. The framework enhances accuracy and provides trustworthy insights into model decisions for military intelligence.
Area of Science:
- Computer Science
- Artificial Intelligence
- Remote Sensing
Background:
- Fine-grained aircraft recognition in remote sensing is crucial for military intelligence.
- Deep learning models lack transparency, limiting their use in high-stakes decisions.
- Existing methods struggle with limited and imbalanced datasets.
Purpose of the Study:
- To develop an explainable aircraft type classification framework (EC-AT).
- To integrate transfer learning, explanation consistency, and trustworthiness evaluation.
- To enhance decision-making transparency in remote sensing applications.
Main Methods:
- Utilized transfer learning with pretrained models for feature representation and efficiency.
- Employed Grad-CAM, LIME, and RISE for complementary explainability analysis.
- Introduced IoMin metric for quantitative explanation consistency assessment.
- Applied Multisource AI Scorecard Table (MAST) for trustworthiness evaluation.
Main Results:
- Transfer learning significantly improved classification accuracy up to 99.9% on limited, imbalanced data.
- Explainability methods identified key aircraft features and potential decision biases.
- Explanation consistency analysis and MAST confirmed framework trustworthiness.
Conclusions:
- The EC-AT framework enhances both performance and explainability for aircraft recognition.
- It offers a trustworthy solution for intelligent remote sensing analysis.
- Provides a valuable reference for high-stakes decision-making in defense intelligence.