Scholarlink

<p>Monitoring wildlife in diverse geographical areas is essential for the conservation of the biological heritage. In this work, we present a study of ospreys <i>Pandion haliaetus</i> nesting in the coast of <i>Todos Los Santos</i> Bay in Ensenada, Baja California. We propose a methodology to automatically examine nests, helping to estimate their levels of visible debris and other contaminants. The used approach is totally non-invasive, fast and cheap. Utilising unmanned aerial vehicles (UAVs), the productivity of six nests was analyzed from 2018 to 2020, yielding an average of 1.5 chicks per nest per year. The recordings and images used in the present study were obtained from a representative sample of those nests. The examination of nests relies on the use of deep learning-based image segmentation and detection in order to expose organic and inorganic materials, e.g. ropes, fishing nets and plastics. We believe that our study could serve as a reference for quickly and reliably assessing the need for nest cleaning when they have more than 50% of non-natural material on the nest surface.</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

Fast assessment of debris in osprey Pandion haliaetus nests using deep learning-based image segmentation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月1日

查看全文
  • R. Daniel Verdugo-Siqueiros,
  • Luis Pellegrin,
  • Gonzalo De León-Girón,
  • Hillary Pérez-Hernández,
  • Jose Angel Gonzalez-Fraga

摘要

Monitoring wildlife in diverse geographical areas is essential for the conservation of the biological heritage. In this work, we present a study of ospreys Pandion haliaetus nesting in the coast of Todos Los Santos Bay in Ensenada, Baja California. We propose a methodology to automatically examine nests, helping to estimate their levels of visible debris and other contaminants. The used approach is totally non-invasive, fast and cheap. Utilising unmanned aerial vehicles (UAVs), the productivity of six nests was analyzed from 2018 to 2020, yielding an average of 1.5 chicks per nest per year. The recordings and images used in the present study were obtained from a representative sample of those nests. The examination of nests relies on the use of deep learning-based image segmentation and detection in order to expose organic and inorganic materials, e.g. ropes, fishing nets and plastics. We believe that our study could serve as a reference for quickly and reliably assessing the need for nest cleaning when they have more than 50% of non-natural material on the nest surface.