Scholarlink

<p>Target detection technology enables real-time monitoring of mud crab growth status, thereby enhancing survival rates and growth efficiency in aquaculture and contributing to the sustainability of breeding systems. This technology reduces environmental impact while supporting eco-friendly resource management, facilitating precise control of aquatic ecosystems. However, due to uneven distribution of mud crabs and the complexity of underwater imaging environments, existing detection methods face significant challenges in improving accuracy. To address these issues, this paper proposes a temporal multi-frame aggregation enhanced detection method based on the YOLO architecture, termed TMAE-YOLO, which incorporates two key innovative modules: the Temporal Multi-frame Aggregation Enhancement (TMAE) module and the Top-Down Progressive Fusion Neck Network (TD-AFPN). The TMAE module enhances detection performance in occluded and blurred scenarios by establishing inter-frame correlations and enabling cross-frame information interaction. The TD-AFPN employs a progressive feature fusion strategy and adaptive spatial fusion technique to ensure smooth transition and efficient integration of features across different levels. This prevents the dissipation or attenuation of high-level semantic features enhanced by TMAE when fused with outputs from other stages, thereby achieving more effective multi-scale feature fusion. The experimental results, conducting on the self-made underwater mud crab dataset, demonstrate that TMAE-YOLO has significant advantages in terms of detection accuracy.</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

TMAE-YOLO: precision detection of mud crabs in underwater environments

  • 期刊论文

  • 发表日期：

    2025年7月25日

查看全文
  • Minghui Yu,
  • Zhijun Xie,
  • Yangfang Ye,
  • Ce Shi

摘要

Target detection technology enables real-time monitoring of mud crab growth status, thereby enhancing survival rates and growth efficiency in aquaculture and contributing to the sustainability of breeding systems. This technology reduces environmental impact while supporting eco-friendly resource management, facilitating precise control of aquatic ecosystems. However, due to uneven distribution of mud crabs and the complexity of underwater imaging environments, existing detection methods face significant challenges in improving accuracy. To address these issues, this paper proposes a temporal multi-frame aggregation enhanced detection method based on the YOLO architecture, termed TMAE-YOLO, which incorporates two key innovative modules: the Temporal Multi-frame Aggregation Enhancement (TMAE) module and the Top-Down Progressive Fusion Neck Network (TD-AFPN). The TMAE module enhances detection performance in occluded and blurred scenarios by establishing inter-frame correlations and enabling cross-frame information interaction. The TD-AFPN employs a progressive feature fusion strategy and adaptive spatial fusion technique to ensure smooth transition and efficient integration of features across different levels. This prevents the dissipation or attenuation of high-level semantic features enhanced by TMAE when fused with outputs from other stages, thereby achieving more effective multi-scale feature fusion. The experimental results, conducting on the self-made underwater mud crab dataset, demonstrate that TMAE-YOLO has significant advantages in terms of detection accuracy.