Scholarlink

<p>Timely and precise identification of plant leaf diseases is essential for efficient crop management, as these diseases significantly affect agricultural productivity and food security. However, previous methods are manual, time-consuming, and often need expert knowledge, which makes them impractical for large-scale farming. Current technologies in Deep Learning (DL) have enabled automatic plant disease detection using image analysis. But differentiating between multiple disease types remains complex due to similarities in visual symptoms, variations in lighting, and data imbalance. So, developing a robust multi-class disease detection model is fundamental to support accurate agriculture and overcome the above issues. Hence, an efficient framework named Wolf Bird Skill Optimization enabled a Deep High Levenberg–Marquardt based Custom Convolutional Network (WBSO_DHLMNet) is designed for multiclass plant disease detection using leaf images. Initially, a midpoint filtering technique is applied to suppress noise while preserving important edge information in leaf images. Next, Region-based Online Selective Examination (ROSE<b>)</b> is employed for effective segmentation of disease-affected regions, improving the focus on relevant features and reducing background interference. Subsequently, data augmentation and feature extraction are performed to enhance dataset diversity and improve feature representation. The classification is carried out using the proposed DHLMNet architecture, which is optimized using the Wolf Bird Skill Optimization (WBSO) algorithm to achieve better convergence, improved exploration–exploitation balance, and enhanced generalization capability. Experimental results demonstrate that WBSO_DHLMNet achieves an accuracy, True Positive Rate (TPR) and False Positive Rate (FPR) is 92.540%, 92.150%, and 6.049%.</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

Wolf Bird Skill Optimization Enabled Deep High Levenberg–Marquardt Based Custom Convolutional Network for Multiclass Plant Disease Detection Using Leaf Images

  • 期刊论文

  • 发表日期：

    2026年6月19日

查看全文
  • Ponnila P,
  • Bazila Banu A

摘要

Timely and precise identification of plant leaf diseases is essential for efficient crop management, as these diseases significantly affect agricultural productivity and food security. However, previous methods are manual, time-consuming, and often need expert knowledge, which makes them impractical for large-scale farming. Current technologies in Deep Learning (DL) have enabled automatic plant disease detection using image analysis. But differentiating between multiple disease types remains complex due to similarities in visual symptoms, variations in lighting, and data imbalance. So, developing a robust multi-class disease detection model is fundamental to support accurate agriculture and overcome the above issues. Hence, an efficient framework named Wolf Bird Skill Optimization enabled a Deep High Levenberg–Marquardt based Custom Convolutional Network (WBSO_DHLMNet) is designed for multiclass plant disease detection using leaf images. Initially, a midpoint filtering technique is applied to suppress noise while preserving important edge information in leaf images. Next, Region-based Online Selective Examination (ROSE) is employed for effective segmentation of disease-affected regions, improving the focus on relevant features and reducing background interference. Subsequently, data augmentation and feature extraction are performed to enhance dataset diversity and improve feature representation. The classification is carried out using the proposed DHLMNet architecture, which is optimized using the Wolf Bird Skill Optimization (WBSO) algorithm to achieve better convergence, improved exploration–exploitation balance, and enhanced generalization capability. Experimental results demonstrate that WBSO_DHLMNet achieves an accuracy, True Positive Rate (TPR) and False Positive Rate (FPR) is 92.540%, 92.150%, and 6.049%.