Scholarlink

<p>Leaf diseases significantly threaten agricultural yields, making their prompt and precise detection crucial for effective management and crop safeguarding. This work aims to create a strong classification framework for detecting leaf diseases in crops by utilizing advanced deep learning methods. The work uses the Dual-Tree Complex Wavelet Preprocessing Framework (DTCWPF) to standardize and reduce noise in input images by maintaining key features for accurate analysis. Feature extraction has been conducted using a blend of EfficientNet-B7 and Inception-ResNet v2 models for capturing a wide range of image attributes such as color, texture, shape, and structural details from both RGB and grayscale images. The Archimedes Optimization Algorithm (AOA) is then applied to select the most pertinent features and to enhance the classification performance.The framework integrates quantum computing principles by employing an Attention-Based Quantum Convolutional Neural Network (ABQCNN), coupled with Adaptive Silver Fox Algorithm (ASFA) for classification. This combination enhances the ability of the model to effectively prioritize salient data features and adaptively optimize feature selection, thereby improving the overall classification performance.To ensure the interpretability of the model predictions, GradCAM and GradCAM + + techniques are applied to generate class-discriminative visual explanations and for highlighting the critical regions within the leaf images those most significantly influence the classification outcomes. These visualization strategies provide valuable insights into the internal decision-making process of the model and subsequently, substantiate its robustness and reliability. The proposed framework has been evaluated on two tomato leaf datasets and two cotton leaf datasets and outstanding performances have been depicted. For the tomato leaf datasets, the model has achieved accuracies of 99.51% and 99.70%, respectively by demonstrating its robustness in classifying various tomato leaf diseases. Similarly, on the cotton leaf datasets, the model has achieved accuracies of 99.92% and 99.60%, respectively by demonstrating its effectiveness in distinguishing different cotton leaf conditions and highlighted the versatility of the framework as well as its potential application in real-world agricultural disease detection scenarios, thereby offering significant advancements in plant pathology.</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

A Quantum-Enhanced Deep Learning Framework for Precise Leaf Disease Detection in Tomato and Cotton Crops

  • 期刊论文

  • 发表日期：

    2025年10月3日

查看全文
  • T. Nivethitha,
  • C. Venkataramanan

摘要

Leaf diseases significantly threaten agricultural yields, making their prompt and precise detection crucial for effective management and crop safeguarding. This work aims to create a strong classification framework for detecting leaf diseases in crops by utilizing advanced deep learning methods. The work uses the Dual-Tree Complex Wavelet Preprocessing Framework (DTCWPF) to standardize and reduce noise in input images by maintaining key features for accurate analysis. Feature extraction has been conducted using a blend of EfficientNet-B7 and Inception-ResNet v2 models for capturing a wide range of image attributes such as color, texture, shape, and structural details from both RGB and grayscale images. The Archimedes Optimization Algorithm (AOA) is then applied to select the most pertinent features and to enhance the classification performance.The framework integrates quantum computing principles by employing an Attention-Based Quantum Convolutional Neural Network (ABQCNN), coupled with Adaptive Silver Fox Algorithm (ASFA) for classification. This combination enhances the ability of the model to effectively prioritize salient data features and adaptively optimize feature selection, thereby improving the overall classification performance.To ensure the interpretability of the model predictions, GradCAM and GradCAM + + techniques are applied to generate class-discriminative visual explanations and for highlighting the critical regions within the leaf images those most significantly influence the classification outcomes. These visualization strategies provide valuable insights into the internal decision-making process of the model and subsequently, substantiate its robustness and reliability. The proposed framework has been evaluated on two tomato leaf datasets and two cotton leaf datasets and outstanding performances have been depicted. For the tomato leaf datasets, the model has achieved accuracies of 99.51% and 99.70%, respectively by demonstrating its robustness in classifying various tomato leaf diseases. Similarly, on the cotton leaf datasets, the model has achieved accuracies of 99.92% and 99.60%, respectively by demonstrating its effectiveness in distinguishing different cotton leaf conditions and highlighted the versatility of the framework as well as its potential application in real-world agricultural disease detection scenarios, thereby offering significant advancements in plant pathology.