Scholarlink

<p>Military camouflaged object detection using unmanned aerial vehicles (UAVs) is a crucial yet challenging task, which can significantly support the intelligence interpretation and accurate strike on the battlefield. However, existing methods primarily focus on the semantic segmentation of the camouflaged object, and neglect the specific needs of military reconnaissance and UAV imagery. To address these limitations, we propose a multi-scale attention and boundary-aware network tailored for military camouflaged object detection from UAV imagery. First, we adopt a pyramid vision transformer as the backbone to extract multi-scale features and produce high-resolution feature maps. Then, we design a feature fusion module to fuse multi-scale features to efficiently transmit feature information. Further, we develop three interdependent modules: boundary extraction module, boundary guidance module, and context fusion module to excavate the boundary and context semantic information, which enhances the feature representation of the military camouflaged object. We also create a new dataset called MCODUAV for military camouflaged object detection from UAV imagery. Extensive experiments on MCODUAV dataset demonstrate that our network has significant advantages in detecting the military camouflaged object from UAV imagery. Our network achieves the highest mean average precision (mAP) by 48.92%, and paves the way for more accurate and preemptive strikes in real-world military operations.</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

Multi-scale attention and boundary-aware network for military camouflaged object detection using unmanned aerial vehicles

  • 期刊论文

  • 发表日期：

    2025年1月3日

查看全文
  • Keshun Liu,
  • Aihua Li,
  • Sen Yang,
  • Changlong Wang,
  • Yuhua Zhang

摘要

Military camouflaged object detection using unmanned aerial vehicles (UAVs) is a crucial yet challenging task, which can significantly support the intelligence interpretation and accurate strike on the battlefield. However, existing methods primarily focus on the semantic segmentation of the camouflaged object, and neglect the specific needs of military reconnaissance and UAV imagery. To address these limitations, we propose a multi-scale attention and boundary-aware network tailored for military camouflaged object detection from UAV imagery. First, we adopt a pyramid vision transformer as the backbone to extract multi-scale features and produce high-resolution feature maps. Then, we design a feature fusion module to fuse multi-scale features to efficiently transmit feature information. Further, we develop three interdependent modules: boundary extraction module, boundary guidance module, and context fusion module to excavate the boundary and context semantic information, which enhances the feature representation of the military camouflaged object. We also create a new dataset called MCODUAV for military camouflaged object detection from UAV imagery. Extensive experiments on MCODUAV dataset demonstrate that our network has significant advantages in detecting the military camouflaged object from UAV imagery. Our network achieves the highest mean average precision (mAP) by 48.92%, and paves the way for more accurate and preemptive strikes in real-world military operations.