测绘通报 ›› 2026, Vol. 0 ›› Issue (9): 108-114.doi: 10.13474/j.cnki.11-2246.2026.0916

• 学术研究 • 上一篇    下一篇

大尺寸无人机影像黄兔尾鼠洞穴识别方法

黄铭琴1, 安佳乐1, 郑江华1, 尼格拉·吐尔逊1, 刘丽1, 王南1, 林峻2, 刘宣1, 徐亚鹏1, 李璇2   

  1. 1. 新疆大学地理与遥感科学学院, 新疆 乌鲁木齐 830046;
    2. 新疆维吾尔自治区草原生物灾害防治中心, 新疆 乌鲁木齐 830001
  • 收稿日期:2025-12-10 发布日期:2026-10-09
  • 通讯作者: 郑江华。E-mail:zheng.jianghua@xju.edu.cn
  • 作者简介:黄铭琴(2000—),女,硕士,主要研究方向为无人机遥感。E-mail:740175476@qq.com
  • 基金资助:
    鼠害遥感专项调查(202434140008);新疆伊犁州直草原生物灾害遥感监测技术应用(202434140015)

An identification method for Eolagurus luteus burrows in large-scale UAV images

Huang Mingqin1, An Jiale1, Zheng Jianghua1, Tuerxun Nigela1, Liu Li1, Wang Nan1, Lin Jun2, Liu Xuan1, Xu Yapeng1, Li Xuan2   

  1. 1. School of Geography and Remote Sensing Science, Xinjiang University, Urumqi 830046, China;
    2. The Center for Grassland Biological Disaster Prevention and Control of Xinjiang, Urumqi 830001, China
  • Received:2025-12-10 Published:2026-10-09

摘要: [目的]针对黄兔尾鼠洞穴因尺寸小、分布广导致人工检测效率低下的问题,本文旨在开发一种基于大尺寸无人机影像的自动化识别方法。[方法]提出一种端到端自动化检测框架,其核心为改进的YOLO-Burrow模型。该模型基于YOLO11n,通过轻量化策略、引入蒙特卡洛注意力机制(MCA)和采用Shape-IoU损失函数进行方法优化。[结果]在自建数据集上的试验表明,模型平均精度(AP)达87%,重组准确率(RA)为97%,模型大小仅0.89 MB,显著优于基准模型。[结论] 本文所提框架可实现大尺寸无人机影像中黄兔尾鼠洞穴的精准自动检测与密度统计,为草原鼠害防控提供可靠技术支撑。

关键词: 无人机, 鼠洞检测, 深度学习, YOLO11n

Abstract: [Purposes]To address the inefficiency of manual detection caused by the small size and wide distribution of Eolagurus luteus burrows,this study aims to develop an automated identification method based on large-scale UAV imagery.[Methods]An end-to-end automated detection framework was proposed,with its core being an improved YOLO-Burrow model.Based on YOLO11n,the model is optimized through a lightweight strategy,the introduction of the Monte Carlo attention (MCA)mechanism,and the adoption of the Shape-IoU loss function.[Findings]Experiments on a self-constructed dataset showed that the model achieved an average precision (AP) of 87%,a reorganized accuracy (RA)of 97%,with a compact model size of only 0.89 MB,significantly outperforming the baseline model.[Conclusions]The proposed framework enables accurate,automatic detection and density assessment of Eolagurus luteus burrows in large-scale UAV images,providing reliable technical support for grassland rodent pest monitoring and control.

Key words: UAV, rodent burrow detection, deep learning, YOLO11n

中图分类号: