Scholarlink

<p>Citrus is one crop with significant added value on a global scale. Early disease diagnosis is crucial for citrus fruits, as well as for all agricultural goods, depending on market needs and potential financial losses. Thus, it is essential to use technology methods to identify citrus illnesses in their early stages, as well as physically damaged citrus fruits. Deep learning techniques have been used to address the challenge of diagnosing diseases of citrus fruits and leaves, as they have lately demonstrated promising results in a number of artificial intelligence applications. This study aims to discover and classify blackspot, canker, greening, and healthy categories, which are frequently encountered in diverse locations. First, many images from the citrus leaves are pre-processed for this purpose. An enhanced Weighted Nuclear Norm Minimization denoising model is used to remove noise during this stage. A special architecture based on an attention-based capsule network is then created. To enhance the ability of the proposed classification, the parameters are fine-tuned using the Extended Osprey Optimization (EOO) algorithm. Using the PlantifyDr dataset, the proposed model achieves 99.17% accuracy, highlighting deep learning’s potential to improve citrus plant disease prediction and classification accuracy.</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

Hierarchical Osprey Residual Attentional Soft Capsule Citrus Model (HORASC2M) for Citrus Plant Disease Classification

  • 期刊论文

  • 发表日期：

    2025年8月7日

查看全文
  • Varsha Santosh Patil,
  • Rina Kamalkumar Bora

摘要

Citrus is one crop with significant added value on a global scale. Early disease diagnosis is crucial for citrus fruits, as well as for all agricultural goods, depending on market needs and potential financial losses. Thus, it is essential to use technology methods to identify citrus illnesses in their early stages, as well as physically damaged citrus fruits. Deep learning techniques have been used to address the challenge of diagnosing diseases of citrus fruits and leaves, as they have lately demonstrated promising results in a number of artificial intelligence applications. This study aims to discover and classify blackspot, canker, greening, and healthy categories, which are frequently encountered in diverse locations. First, many images from the citrus leaves are pre-processed for this purpose. An enhanced Weighted Nuclear Norm Minimization denoising model is used to remove noise during this stage. A special architecture based on an attention-based capsule network is then created. To enhance the ability of the proposed classification, the parameters are fine-tuned using the Extended Osprey Optimization (EOO) algorithm. Using the PlantifyDr dataset, the proposed model achieves 99.17% accuracy, highlighting deep learning’s potential to improve citrus plant disease prediction and classification accuracy.