Scholarlink

<p>Marine object detection in underwater environments faces significant challenges due to degraded image quality, poor visibility, and color distortion. To tackle these issues, we introduce MarineDetNet, a deep CNN-based image enhancement network designed specifically for underwater imagery. MarineDetNet employs custom loss functions to address color, contrast, illumination, and exposure issues, thereby enhancing image quality and subsequently improving the accuracy of marine object detection. By integrating MarineDetNet with YOLOv8, we demonstrate improved performance, achieving a mean average precision (mAP) of 80.2% on the DUO dataset, outperforming non-enhanced detection (72% mAP). Our approach offers a lightweight, cost-effective solution for real-time marine object detection, with potential applications in autonomous underwater vehicles and robotics. This work highlights the importance of image enhancement for enhancing the effectiveness of underwater object detection.</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

MarineDetNet: Enhancing Underwater Image Quality for Improved Marine Object Detection

  • 期刊论文

  • 发表日期：

    2025年6月27日

查看全文
  • B Lakshmanan,
  • B Selvakumar,
  • R Nivetika,
  • R Karpagalakshmi

摘要

Marine object detection in underwater environments faces significant challenges due to degraded image quality, poor visibility, and color distortion. To tackle these issues, we introduce MarineDetNet, a deep CNN-based image enhancement network designed specifically for underwater imagery. MarineDetNet employs custom loss functions to address color, contrast, illumination, and exposure issues, thereby enhancing image quality and subsequently improving the accuracy of marine object detection. By integrating MarineDetNet with YOLOv8, we demonstrate improved performance, achieving a mean average precision (mAP) of 80.2% on the DUO dataset, outperforming non-enhanced detection (72% mAP). Our approach offers a lightweight, cost-effective solution for real-time marine object detection, with potential applications in autonomous underwater vehicles and robotics. This work highlights the importance of image enhancement for enhancing the effectiveness of underwater object detection.