Scholarlink

<p>This study explores the application of machine learning techniques for the automated identification of <i>Eucalyptus</i> trees based on bark images. The models were trained on a dataset containing images of <i>Eucalyptus</i> and non-Eucalyptus trees, with the Eucalyptus samples collected from Rmilat Park in Tangier, Morocco—a location not previously studied for this purpose, underscoring the originality of this research. The methodology involves data collection, image preprocessing, feature extraction, and classification using support vector machine (<i>SVM)</i> and K-nearest neighbors <i>(KNN)</i>. Three types of features were investigated: <i>raw pixel intensity values, histogram of oriented gradients (HOG), and local binary patterns (LBP).</i> While pixel intensities provided a simple baseline, HOG captured edge and shape information, and LBP encoded local texture patterns. Among the models, SVM using HOG features achieved the highest accuracy at 88.31%, followed by KNN using LBP features at 86.36%, while the model using <i>raw pixel values</i> performed the least effectively, with an accuracy of <i>84%</i>. These results demonstrate the effectiveness of bark-based identification as a reliable and non-invasive method for tree species classification. The approach is particularly valuable in scenarios where leaves or flowers are absent, such as during dry seasons or in degraded environments.</p><p>Furthermore, this technique has practical implications for remote sensing and forest monitoring, offering a cost-effective alternative to traditional field surveys. It also supports the development of mobile applications or decision support systems for foresters, enabling real-time, in-field species identification and contributing to improved forest management and biodiversity conservation.</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

Comparing support vector machines and k-nearest neighbors for Eucalyptus tree identification using bark images

  • 期刊论文

  • 发表日期：

    2025年9月19日

查看全文
  • Akram Elidrissi,
  • Ibrahim Hadj Baraka,
  • Abdelali Astito,
  • abdes-samed Bernoussi

摘要

This study explores the application of machine learning techniques for the automated identification of Eucalyptus trees based on bark images. The models were trained on a dataset containing images of Eucalyptus and non-Eucalyptus trees, with the Eucalyptus samples collected from Rmilat Park in Tangier, Morocco—a location not previously studied for this purpose, underscoring the originality of this research. The methodology involves data collection, image preprocessing, feature extraction, and classification using support vector machine (SVM) and K-nearest neighbors (KNN). Three types of features were investigated: raw pixel intensity values, histogram of oriented gradients (HOG), and local binary patterns (LBP). While pixel intensities provided a simple baseline, HOG captured edge and shape information, and LBP encoded local texture patterns. Among the models, SVM using HOG features achieved the highest accuracy at 88.31%, followed by KNN using LBP features at 86.36%, while the model using raw pixel values performed the least effectively, with an accuracy of 84%. These results demonstrate the effectiveness of bark-based identification as a reliable and non-invasive method for tree species classification. The approach is particularly valuable in scenarios where leaves or flowers are absent, such as during dry seasons or in degraded environments.

Furthermore, this technique has practical implications for remote sensing and forest monitoring, offering a cost-effective alternative to traditional field surveys. It also supports the development of mobile applications or decision support systems for foresters, enabling real-time, in-field species identification and contributing to improved forest management and biodiversity conservation.