Scholarlink

<p>In response to the urgent need for water environment protection, this study proposes an improved algorithm for detecting floating objects on the surface of water: You Only Look Once version 8-water surface floating object detection (YOLOv8-WSFOD). This algorithm aims to address the impacts of illumination variations and water surface distortion on floating object detection, as well as missed and false small object detections in complex aquatic scenarios. The improvements provided by YOLOv8-WSFOD include the design of a spatial pyramid pooling fusion-large separable kernel attention (SPPF-LSKA) module to evolve the SPPF module, which possesses long-range dependence and adaptive capabilities to mitigate the interference caused by noise factors such as water surface fluctuations, strong illumination, and high-contrast weather conditions. Then, the model focuses more on important feature areas and effectively reduces the impact of noise on floating object detection. Additionally, the normalized Wasserstein distance (NWD) regression loss function is introduced and combined with the original complete intersection-over-union (CIoU) loss function through weighted integration, resulting in a novel comprehensive regression loss function that significantly enhances the small object detection performance and accuracy of the model. Finally, the adaptive moment estimation (Adam) optimizer in the original algorithm was replaced with the second-order clipped stochastic optimization (Sophia) optimizer to improve the generalizability of the model. Experimental results demonstrate that YOLOv8-WSFOD achieves significant WSFOD improvements, with a 3.4% mean average precision (mAP)@0.5 increase and a 3.4% mAP@0.5:0.95 increase. In multiclass detection tasks, mAP@0.5 improves by 8.6%, and the detection accuracies achieved across all categories is significantly enhanced. The performance of the developed algorithm demonstrates its great feasibility for deployment in unmanned cleaning vessels, offering high accuracy and efficiency in WSFOD scenarios.</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

Research on the intelligent detection and analysis of floating debris in polluted water

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月20日

查看全文
  • Huizhen Dong,
  • Jianjun Li,
  • Bei Li

摘要

In response to the urgent need for water environment protection, this study proposes an improved algorithm for detecting floating objects on the surface of water: You Only Look Once version 8-water surface floating object detection (YOLOv8-WSFOD). This algorithm aims to address the impacts of illumination variations and water surface distortion on floating object detection, as well as missed and false small object detections in complex aquatic scenarios. The improvements provided by YOLOv8-WSFOD include the design of a spatial pyramid pooling fusion-large separable kernel attention (SPPF-LSKA) module to evolve the SPPF module, which possesses long-range dependence and adaptive capabilities to mitigate the interference caused by noise factors such as water surface fluctuations, strong illumination, and high-contrast weather conditions. Then, the model focuses more on important feature areas and effectively reduces the impact of noise on floating object detection. Additionally, the normalized Wasserstein distance (NWD) regression loss function is introduced and combined with the original complete intersection-over-union (CIoU) loss function through weighted integration, resulting in a novel comprehensive regression loss function that significantly enhances the small object detection performance and accuracy of the model. Finally, the adaptive moment estimation (Adam) optimizer in the original algorithm was replaced with the second-order clipped stochastic optimization (Sophia) optimizer to improve the generalizability of the model. Experimental results demonstrate that YOLOv8-WSFOD achieves significant WSFOD improvements, with a 3.4% mean average precision (mAP)@0.5 increase and a 3.4% mAP@0.5:0.95 increase. In multiclass detection tasks, mAP@0.5 improves by 8.6%, and the detection accuracies achieved across all categories is significantly enhanced. The performance of the developed algorithm demonstrates its great feasibility for deployment in unmanned cleaning vessels, offering high accuracy and efficiency in WSFOD scenarios.