Scholarlink

<p>In response to the challenges posed by large-scale variations and the difficulty in extracting features of small objects in unmanned aerial vehicle (UAV) images, this paper proposes a novel small object detection method based on YOLOv8. Firstly, a novel efficient multi-scale attention feature pyramid network is designed, which effectively enhances information representation across different scales in deep features while reducing computational overhead. Secondly, a receptive field enhancement module is proposed, with the purpose of expanding the receptive field and enabling the extraction of rich and valuable multiscale feature information. Finally, the approach integrates a tiny object detection layer into the head network with the objective of capturing tiny objects in UAV images, thereby achieving a significant reduction in the missed detection rate and improvement in detection accuracy. Experiments on the VisDrone and UAVDT datasets demonstrate that the proposed method achieves higher precision in small object detection and outperforms existing methods in detection performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A UAV image small object detection approach with enhanced feature pyramid and receptive field

  • 期刊论文

  • 发表日期：

    2025年9月11日

查看全文
  • Huaping Zhou,
  • Zhiyong Sun,
  • Kelei Sun

摘要

In response to the challenges posed by large-scale variations and the difficulty in extracting features of small objects in unmanned aerial vehicle (UAV) images, this paper proposes a novel small object detection method based on YOLOv8. Firstly, a novel efficient multi-scale attention feature pyramid network is designed, which effectively enhances information representation across different scales in deep features while reducing computational overhead. Secondly, a receptive field enhancement module is proposed, with the purpose of expanding the receptive field and enabling the extraction of rich and valuable multiscale feature information. Finally, the approach integrates a tiny object detection layer into the head network with the objective of capturing tiny objects in UAV images, thereby achieving a significant reduction in the missed detection rate and improvement in detection accuracy. Experiments on the VisDrone and UAVDT datasets demonstrate that the proposed method achieves higher precision in small object detection and outperforms existing methods in detection performance.