Scholarlink

<p>Advancements in agriculture require precise and efficient plant growth monitoring techniques, particularly in vertical farming systems. While deep learning models have demonstrated potential for real-time detection of plant diseases, their effectiveness in plant growth detection remains underexplored. This study uses a custom-annotated basil plant dataset to evaluate the performance of YOLOv8, YOLOv9, and YOLOv11 in detecting growth stages—harvesting, nursery, and young—under dynamic environmental conditions. Model performance was assessed using precision, recall, and mean Average Precision (mAP) at intersection over a union threshold of 0.50 (mAP50), where YOLOv8 achieved an overall mAP of 0.842, with harvesting performing best (recall: 0.924) and nursery the lowest (recall: 0.462). YOLOv9 improved with a mAP of 0.865, a harvesting recall of 0.953, and a nursery recall of 0.731. While YOLOv11 offers similar performance with a mAP of 0.846 and excels in harvesting recall (0.917), nursery performance remained lower (recall: 0.727). Among the models, YOLOv9 provided the best balance between accuracy and reduced misclassifications for all classes, making it the most suitable for real-time vertical farming applications. These findings highlight the strengths and limitations of deep learning-based plant monitoring, advancing sustainable agriculture.</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

Enhancing precision farming with YOLO models for real-time growth detection of basil plants

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • Ame Saleh Ali,
  • Jyotindra Narayan,
  • Ashwani Shukla,
  • Raj Shree

摘要

Advancements in agriculture require precise and efficient plant growth monitoring techniques, particularly in vertical farming systems. While deep learning models have demonstrated potential for real-time detection of plant diseases, their effectiveness in plant growth detection remains underexplored. This study uses a custom-annotated basil plant dataset to evaluate the performance of YOLOv8, YOLOv9, and YOLOv11 in detecting growth stages—harvesting, nursery, and young—under dynamic environmental conditions. Model performance was assessed using precision, recall, and mean Average Precision (mAP) at intersection over a union threshold of 0.50 (mAP50), where YOLOv8 achieved an overall mAP of 0.842, with harvesting performing best (recall: 0.924) and nursery the lowest (recall: 0.462). YOLOv9 improved with a mAP of 0.865, a harvesting recall of 0.953, and a nursery recall of 0.731. While YOLOv11 offers similar performance with a mAP of 0.846 and excels in harvesting recall (0.917), nursery performance remained lower (recall: 0.727). Among the models, YOLOv9 provided the best balance between accuracy and reduced misclassifications for all classes, making it the most suitable for real-time vertical farming applications. These findings highlight the strengths and limitations of deep learning-based plant monitoring, advancing sustainable agriculture.