Related Experiment Video
Updated: Jul 19, 2025
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
1.5K
Explicable Fine-Grained Aircraft Recognition Via Deep Part Parsing Prior Framework for High-Resolution Remote Sensing
IEEE Transactions on Cybernetics
|August 8, 2023
Summary
This study introduces APPEAR, a knowledge-driven deep learning method for aircraft recognition using part parsing priors. It improves feature extraction and recognition performance, especially with limited data.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Remote Sensing
Background:
- Aircraft recognition is vital for civil and military applications, with high-spatial resolution remote sensing being a key approach.
- Current data-driven methods struggle with limited training data, hindering effective feature extraction and leading to suboptimal recognition performance.
- Existing methods lack the ability to precisely locate discriminative regions within aircraft imagery.
Purpose of the Study:
- To develop a novel knowledge-driven deep learning framework for explicable aircraft recognition.
- To address the limitations of data-driven methods in feature extraction and recognition performance, particularly with scarce training data.
- To enhance aircraft recognition by explicitly modeling aircraft structure and leveraging part-based information.
Main Methods:
- Proposed the explicable aircraft recognition framework based on a part parsing prior (APPEAR).
- Modeled aircraft structure using a pixel-level part parsing prior, dividing aircraft into five key parts (nose, wings, fuselage, tail).
- Introduced a knowledge-driven aircraft part attention (KAPA) module for geometric-invariant feature representation and adaptive reweighting of part importance.
Main Results:
- The APPEAR framework achieved superior performance on two aircraft recognition datasets.
- Demonstrated robustness in few-shot learning scenarios, indicating effectiveness with limited data.
- Ablation studies identified the fuselage and wings as the most discriminative parts for aircraft recognition.
Conclusions:
- The proposed APPEAR framework effectively utilizes part parsing priors to improve aircraft recognition accuracy.
- The knowledge-driven approach, particularly the KAPA module, enhances feature extraction by focusing on discriminative parts.
- APPEAR offers a robust solution for aircraft recognition, outperforming existing methods and showing promise for few-shot learning applications.
