Scholarlink

<p>This study explores the efficacy of Pléiades data for distinguishing different morphological classes using the feature extraction method. Recent advances in remote sensing have made it possible to use extracted features coupled with Artificial Intelligence (AI) and Deep Learning (DL) to achieve better accuracy. This paper deals with feature extraction by Grey Level Co-occurrence Matrix (GLCM) and four classification techniques, namely Support Vector Machine (SVM), Random Forest (RF), Convolutional Neural Network (CNN), and Classification and Regression Tree (CART). The highest classification accuracy for a scene of Pléiades data at Sai Yok national park, Thailand, was obtained for SVM, followed by CNN and RF. In contrast, the lowest accuracy was observed for the CART algorithm. Our results suggest that feature extraction has drastically improved accuracy for all four classification techniques, and the improvement is substantial (for SVM, 74.34–91.8%; for RF, 70.13–90.23%; for CNN, 61.45–89.45%; for CART, 54.02–84.62%). However, the classification algorithms performed differently in separating closely aligned classes, like cropland and forest, or road and built-up areas. The proximity of the two separate groups of features in the feature space image further proves the complexity of the problem. The separate distribution of the pixels in the feature space suggests the efficiency of feature extraction. Additionally, because of the high accuracy levels (&gt; 89% overall accuracy for SVM, RF, and CNN), this research suggests using Pléiades data for land use analysis in the future.</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

Using Artificial Intelligence and Deep Learning Algorithms to Extract Land Features from High-Resolution Pléiades Data

  • 期刊论文

  • 发表日期：

    2025年1月2日

查看全文
  • Anirban Mukhopadhyay,
  • Indrajit Pal,
  • Niloy Pramanick,
  • Rituparna Acharyya,
  • Jyoti Prakash Hati,
  • Sudipta Mukherjee,
  • Ganni S. V. S. Aditya Bharadwaz,
  • Debasish Mitra

摘要

This study explores the efficacy of Pléiades data for distinguishing different morphological classes using the feature extraction method. Recent advances in remote sensing have made it possible to use extracted features coupled with Artificial Intelligence (AI) and Deep Learning (DL) to achieve better accuracy. This paper deals with feature extraction by Grey Level Co-occurrence Matrix (GLCM) and four classification techniques, namely Support Vector Machine (SVM), Random Forest (RF), Convolutional Neural Network (CNN), and Classification and Regression Tree (CART). The highest classification accuracy for a scene of Pléiades data at Sai Yok national park, Thailand, was obtained for SVM, followed by CNN and RF. In contrast, the lowest accuracy was observed for the CART algorithm. Our results suggest that feature extraction has drastically improved accuracy for all four classification techniques, and the improvement is substantial (for SVM, 74.34–91.8%; for RF, 70.13–90.23%; for CNN, 61.45–89.45%; for CART, 54.02–84.62%). However, the classification algorithms performed differently in separating closely aligned classes, like cropland and forest, or road and built-up areas. The proximity of the two separate groups of features in the feature space image further proves the complexity of the problem. The separate distribution of the pixels in the feature space suggests the efficiency of feature extraction. Additionally, because of the high accuracy levels (> 89% overall accuracy for SVM, RF, and CNN), this research suggests using Pléiades data for land use analysis in the future.