Scholarlink

<p>Controlling diseases on the foliage of cultivated plants has a&#xa0;major impact on crop yield and quality. Leaf symptoms and environmental information are the basis of crop disease and pest identification. Due to the complexity and diversity of disease leaf symptoms and environmental information, the accuracy of crop disease detection methods has decreased. Aiming at low disease identification efficiency and difficult feature modeling, we propose a&#xa0;new deep learning algorithm and obtain ideal results. We used improved convolutional neural networks (CNNs) SPDNet and GrNet (SPDGrNet) to perform <i>Zea mays</i> disease identification. The proposed research and analysis focus on four categories of <i>Zea mays</i> symptoms: healthy, leaf blight, southern leaf blight and grey spot. The experimental results and analysis confirm that the proposed approach outperforms other excellent plant disease recognition algorithms and achieves an overall classification accuracy of 98.96%. When compared to AlexNet, VGG-16, Inception-V4, ShuffleNet-V3, and Efficient-B7, the proposed model achieves accuracy improvements of 27.78%, 19.05%, 14.59%, 10.28%, and 7.06%, respectively, with corresponding accuracies of 71.18%, 79.91%, 84.37%, 88.68%, and 91.9%. The proposed model has outstanding generalization ability in open-source databases, such as the PlantifyDr, Plant Pathology, and New Plant Diseases Dataset. The suggested strategy can provide ideas for transplanting models into mobile disease detection equipment. This is very helpful in the intelligent development of agricultural systems and crop leaf disease analysis.</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

SPDGrNet: A Lightweight and Efficient Image Classification Network for Zea mays Diseases

  • 期刊论文

  • 发表日期：

    2025年5月16日

查看全文
  • Laixiang Xu,
  • Bingxu Cao,
  • Jing feng Su,
  • Fengyue Zhang,
  • Hongyun Song,
  • Junmin Zhao

摘要

Controlling diseases on the foliage of cultivated plants has a major impact on crop yield and quality. Leaf symptoms and environmental information are the basis of crop disease and pest identification. Due to the complexity and diversity of disease leaf symptoms and environmental information, the accuracy of crop disease detection methods has decreased. Aiming at low disease identification efficiency and difficult feature modeling, we propose a new deep learning algorithm and obtain ideal results. We used improved convolutional neural networks (CNNs) SPDNet and GrNet (SPDGrNet) to perform Zea mays disease identification. The proposed research and analysis focus on four categories of Zea mays symptoms: healthy, leaf blight, southern leaf blight and grey spot. The experimental results and analysis confirm that the proposed approach outperforms other excellent plant disease recognition algorithms and achieves an overall classification accuracy of 98.96%. When compared to AlexNet, VGG-16, Inception-V4, ShuffleNet-V3, and Efficient-B7, the proposed model achieves accuracy improvements of 27.78%, 19.05%, 14.59%, 10.28%, and 7.06%, respectively, with corresponding accuracies of 71.18%, 79.91%, 84.37%, 88.68%, and 91.9%. The proposed model has outstanding generalization ability in open-source databases, such as the PlantifyDr, Plant Pathology, and New Plant Diseases Dataset. The suggested strategy can provide ideas for transplanting models into mobile disease detection equipment. This is very helpful in the intelligent development of agricultural systems and crop leaf disease analysis.