跳到主要导航 跳到搜索 跳到主要内容

YOLO-AFP: A More Robust Network for Aerial Object Detection

  • Xue Li
  • , Ziang Wang*
  • , Xueyu Chen
  • , Wangbin Li
  • , Kaimin Sun
  • , Zuomei Lai
  • , Peipei Zhu
  • *此作品的通讯作者
  • Southwest China Inst Elect Technol
  • Natl Key Lab Complex Aviat Syst Simulat

科研成果: 期刊稿件 › 论文 › 同行评审

摘要

In practical applications of aerial object detection, real-time uncrewed aerial vehicle (UAV) imagery is often affected by noise, low light, and cloud occlusion, leading to poor image quality. The performance of mainstream UAV object detection algorithms tends to degrade when applied to such imagery, as these models are typically trained and evaluated on clean datasets. To address these challenges, we propose a robust YOLO-based network, YOLO-atrous feature pyramid (AFP), which integrates an AFP module. This allows the model to generalize effectively under various corrupted conditions, despite being trained only on clean data. First, we introduced AFP module, which employs atrous convolutions with varying dilation rates, and integrate it into the path aggregation network to expand the receptive field. This enhancement allows the model to better capture object-background relationships and reduce feature corruption caused by local pixel changes. Second, we propose a robust ResNet-spatial pyramid pooling fast (SPPF) as the backbone network, which retains strong feature extraction capabilities while having fewer residual connections compared to YOLO's Darknet. This design effectively mitigates the impact of corrupted image features on subsequent feature extraction. To validate the effectiveness of our method, we constructed two new datasets based on the DOTAv1.0 dataset, named DOTA-HC and DOTA-HCloud. Experimental results demonstrate that on the DOTA-HC dataset, YOLO-AFP achieved an mean performance under corruption of 60.8% and an relative performance under corruption (rPC) of 80.3%, outperforming the best real-time detection model by 1.5% and 2%, respectively. On the DOTA-HCloud dataset, YOLO-AFP achieved an rPC of 88.5%, surpassing the top model by 1.1% .
源语言英语
页(从-至)24466-24478
页数13
期刊IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
卷18
DOI
出版状态已出版 - 1月 2025

基金资助

This work was supported in part by the National Key Research and Development Program of China under Grant 2022YFB3902900, in part by the National Natural Science Foundation of China under Grant 62303433, and in part by the Shanghai Aerospace Technology Innovation Fund under Grant SAST2023-046.

关键词

  • Object detection
  • Feature extraction
  • Robustness
  • Remote sensing
  • Real-time systems
  • Accuracy
  • Data models
  • Detectors
  • Training
  • YOLO
  • Aerial images
  • atrous convolution
  • object detection
  • robustness

收录

  • EI
  • SCIE
  • Scopus

学术指纹

探究 'YOLO-AFP: A More Robust Network for Aerial Object Detection' 的科研主题。它们共同构成独一无二的学术指纹。

引用此