Scholarlink

<p>Technological advancements have revolutionized agriculture, particularly in early and accurate crop disease detection, which is crucial for improving crop yields. Image-based machine learning has emerged as a critical approach for identifying and classifying plant diseases, with success relying on the extraction of discriminative features from crop images. This study analyzed 5,932 field images of rice leaves affected by bacterial blight, blast, brown spot, and tungro. The images were standardized to 300 × 300 pixels and processed to generate image embeddings, commonly referred to as deep features, using pre-trained models such as Inception V3, VGG16, VGG19, SqueezeNet, and Deeploc. These features were classified using Support Vector Machine (SVM) kernels, with the radial basis function (RBF) kernel delivering superior performance. Specifically, the SVM-RBF kernel achieved an AUC of 100.00, classification accuracy (CA) of 99.999%, F1 score of 99.998%, precision of 99.998%, and recall of 99.997% when combined with Inception V3. These results highlight the effectiveness of combining advanced image-embedding techniques with SVM classifiers for precise rice leaf disease classification, offering a robust methodology to enhance agricultural productivity.</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

Enhancing rice leaf disease classification accuracy with svm kernels using image embedding techniques for feature extraction: a comprehensive analysis

  • 期刊论文

  • 发表日期：

    2025年4月30日

查看全文
  • S. Ratan Kumar,
  • M. Appala Srinuvasu,
  • A. Vamsidhar,
  • B. Dinesh Reddy

摘要

Technological advancements have revolutionized agriculture, particularly in early and accurate crop disease detection, which is crucial for improving crop yields. Image-based machine learning has emerged as a critical approach for identifying and classifying plant diseases, with success relying on the extraction of discriminative features from crop images. This study analyzed 5,932 field images of rice leaves affected by bacterial blight, blast, brown spot, and tungro. The images were standardized to 300 × 300 pixels and processed to generate image embeddings, commonly referred to as deep features, using pre-trained models such as Inception V3, VGG16, VGG19, SqueezeNet, and Deeploc. These features were classified using Support Vector Machine (SVM) kernels, with the radial basis function (RBF) kernel delivering superior performance. Specifically, the SVM-RBF kernel achieved an AUC of 100.00, classification accuracy (CA) of 99.999%, F1 score of 99.998%, precision of 99.998%, and recall of 99.997% when combined with Inception V3. These results highlight the effectiveness of combining advanced image-embedding techniques with SVM classifiers for precise rice leaf disease classification, offering a robust methodology to enhance agricultural productivity.