基于超分辨率重构与改进YOLOv8的无人机捕获图像中蟹钉检测技术
计算机视觉与模式识别
2025-07-09 v2
摘要
蟹钉在 coastal ecosystems 中发挥着重要作用，包括散播种子、清理碎屑和扰动土壤，作为海洋环境健康的重要指示物。传统调查方法（如四分图抽样）代价高昂、耗时且受环境依赖。本研究提出一种创新方法，将无人机遥感与超分辨率重构（Super-Resolution Reconstruction, SRR）以及CRAB-YOLO检测网络相结合，后者是YOLOv8s的改进版本，以监测蟹钉。SRR通过解决运动模糊和分辨率不足的问题，显著提高了检测准确度。CRAB-YOLO网络集成了三项改进，针对检测准确度、蟹钉特征和计算效率，实现了各类主流检测模型的SOTA性能。RDN网络表现最佳图像重建，CRAB-YOLO在SRR测试集上实现69.5%的平均精度（mAP），相较于常规Bicubic方法提升40%。这些结果表明，所提方法有效检测蟹钉，为大规模蟹钉监测提供了成本效益高的自动化解决方案，有助于 coastal benthos 保护。
引用
@article{arxiv.2408.03559,
title = {Enhanced hermit crabs detection using super-resolution reconstruction and improved YOLOv8 on UAV-captured imagery},
author = {Fan Zhao and Yijia Chen and Dianhan Xi and Yongying Liu and Jiaqi Wang and Shigeru Tabeta and Katsunori Mizuno},
journal= {arXiv preprint arXiv:2408.03559},
year = {2025}
}
备注
The earlier version of this conference paper was presented at OCEANS 2024-Singapore and was selected for inclusion in the Student Poster Competition (SPC) Program, the final version of this project was published in the academic journal Marine Environmental Research with the Doi: https://doi.org/10.1016/j.marenvres.2025.107313