Scholarlink

<p>Invasion of the golden mussel (<i>Limnoperna fortunei</i>) into Brazilian watersheds is still impacting energy production and aquaculture after 30&#xa0;years of establishment. No control attempts have been effective and even monitoring is limited by costs and accessibility to the areas. In this context, we propose an approach that integrates traditional monitoring tools with an artificial intelligence (AI) program developed using convolutional neural networks (CNN), with the aim to identify and quantify golden mussels in two Brazilian watersheds, Paraná and São Francisco. In the latter, we conducted an additional 7-month temporal evaluation using recruitment plates. Neural networks can assist in species identification in complex environments, facilitating population monitoring and biodiversity assessment. In our study, the AI program had 85–98% accuracy compared to human measurements, demonstrating a high success rate for autonomous assessment. Independent of individual mussel size, the best feature for golden mussel detection was valve aperture. This study provides a valuable quantitative and methodological baseline regarding golden mussel infestation in Brazil, highlighting the efficiency and cost-effectiveness of the AI program for monitoring this invasive species. This technique is easily replicable and scalable, with a great potential to facilitate the work of researchers and environmental agencies worldwide as an additional tool to combat the spread of invasive species.</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

Artificial intelligence as an ally to assess and manage the golden mussel (Limnoperna fortunei (Dunker, 1857)) bioinvasion

  • 期刊论文

  • 发表日期：

    2025年3月24日

查看全文
  • Humberto F. M. Fortunato,
  • Raquel M. A. Figueira,
  • Ronny F. M. de Souza,
  • Nelson Theodoro Junior,
  • Victor B. B. Mello

摘要

Invasion of the golden mussel (Limnoperna fortunei) into Brazilian watersheds is still impacting energy production and aquaculture after 30 years of establishment. No control attempts have been effective and even monitoring is limited by costs and accessibility to the areas. In this context, we propose an approach that integrates traditional monitoring tools with an artificial intelligence (AI) program developed using convolutional neural networks (CNN), with the aim to identify and quantify golden mussels in two Brazilian watersheds, Paraná and São Francisco. In the latter, we conducted an additional 7-month temporal evaluation using recruitment plates. Neural networks can assist in species identification in complex environments, facilitating population monitoring and biodiversity assessment. In our study, the AI program had 85–98% accuracy compared to human measurements, demonstrating a high success rate for autonomous assessment. Independent of individual mussel size, the best feature for golden mussel detection was valve aperture. This study provides a valuable quantitative and methodological baseline regarding golden mussel infestation in Brazil, highlighting the efficiency and cost-effectiveness of the AI program for monitoring this invasive species. This technique is easily replicable and scalable, with a great potential to facilitate the work of researchers and environmental agencies worldwide as an additional tool to combat the spread of invasive species.