Scholarlink

<p>A lightweight remote sensing image ship detection model based on YOLOv11n, YOLOv11 ShipNet, is proposed to address the technical challenges faced by ship detection in complex ocean remote sensing scenarios, such as severe background interference, high missed detection rate of small targets, and uneven sample regression. The model partially replaces the original C3k2 structure with C2f modules with cross layer skip connections, effectively enhancing the network’s gradient conduction efficiency and feature reuse capability; Design a dual channel ECA Attention Plus attention mechanism to enhance the model’s ability to focus and perceive key target areas on ships; The Focal MPDIOU loss function based on IoU threshold dynamic modulation is adopted to implement adaptive weight attenuation on high IoU prone regression samples, significantly improving the accuracy of small object detection and the convergence speed of bounding box regression. The experimental results on the HRSC2016 high-resolution remote sensing ship dataset show that the proposed algorithm mAP@0.5 Reaching 91.3%, which is 1.5 percentage points higher than the benchmark YOLOv11n, the performance is significantly better than other mainstream object detection models, fully verifying the effectiveness and superiority of this method.</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

YOLOv11-ShipNet: a remote sensing image ship detection model based on dual-domain attention and focal-MPDIoU loss

  • 期刊论文

  • 发表日期：

    2026年7月14日

查看全文
  • Han Zhang,
  • Peixue Liu,
  • Yujie Chen,
  • Shu Liu,
  • Zexin Li

摘要

A lightweight remote sensing image ship detection model based on YOLOv11n, YOLOv11 ShipNet, is proposed to address the technical challenges faced by ship detection in complex ocean remote sensing scenarios, such as severe background interference, high missed detection rate of small targets, and uneven sample regression. The model partially replaces the original C3k2 structure with C2f modules with cross layer skip connections, effectively enhancing the network’s gradient conduction efficiency and feature reuse capability; Design a dual channel ECA Attention Plus attention mechanism to enhance the model’s ability to focus and perceive key target areas on ships; The Focal MPDIOU loss function based on IoU threshold dynamic modulation is adopted to implement adaptive weight attenuation on high IoU prone regression samples, significantly improving the accuracy of small object detection and the convergence speed of bounding box regression. The experimental results on the HRSC2016 high-resolution remote sensing ship dataset show that the proposed algorithm mAP@0.5 Reaching 91.3%, which is 1.5 percentage points higher than the benchmark YOLOv11n, the performance is significantly better than other mainstream object detection models, fully verifying the effectiveness and superiority of this method.