Scholarlink

<p>Early and accurate detection of crop diseases is vital for improving yield and ensuring food security. Strawberry plants, in particular, are vulnerable to various leaf diseases that affect their productivity. This paper introduces the Hierarchical Feature Extraction Super-Resolution Network (HFESRN), a novel deep learning framework tailored to enhance Low-Resolution (LR) images of strawberry leaves for disease diagnosis. HFESRN integrates two innovations: the Diverse Scale Group Convolutional Network (DSGCN), which simultaneously captures multi-scale spatial and channel features while preserving fine-grained details, and the Feature Segregation and Aggregation Block (FSAB), which combines multiple DSGCNs with a Channel Aggregation Module (CAM) to progressively refine feature representations. The super-resolved images are then classified using a lightweight CNN-based Disease Detection Network (DDN). HFESRN achieves high PSNR and SSIM values while significantly boosting classification accuracy. On the Strawberry Leaf Disease dataset, the proposed model achieved accuracies of 99.37%, 98.73%, and 97.47% for 2 ×, 4 ×, and 6 × scaling factors, respectively, outperforming state-of-the-art methods. This model offers a practical solution for precision agriculture by enabling reliable disease detection from low-resolution images.</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

Combating strawberry leaf diseases with image super resolution vision

  • 期刊论文

  • 发表日期：

    2025年7月29日

查看全文
  • P. V. Yeswanth,
  • S. Deivalakshmi

摘要

Early and accurate detection of crop diseases is vital for improving yield and ensuring food security. Strawberry plants, in particular, are vulnerable to various leaf diseases that affect their productivity. This paper introduces the Hierarchical Feature Extraction Super-Resolution Network (HFESRN), a novel deep learning framework tailored to enhance Low-Resolution (LR) images of strawberry leaves for disease diagnosis. HFESRN integrates two innovations: the Diverse Scale Group Convolutional Network (DSGCN), which simultaneously captures multi-scale spatial and channel features while preserving fine-grained details, and the Feature Segregation and Aggregation Block (FSAB), which combines multiple DSGCNs with a Channel Aggregation Module (CAM) to progressively refine feature representations. The super-resolved images are then classified using a lightweight CNN-based Disease Detection Network (DDN). HFESRN achieves high PSNR and SSIM values while significantly boosting classification accuracy. On the Strawberry Leaf Disease dataset, the proposed model achieved accuracies of 99.37%, 98.73%, and 97.47% for 2 ×, 4 ×, and 6 × scaling factors, respectively, outperforming state-of-the-art methods. This model offers a practical solution for precision agriculture by enabling reliable disease detection from low-resolution images.