Scholarlink

<p>Hyperspectral remote sensing has evolved as an effective method for mineral identification and mapping, delivering comprehensive spectral data over a wide range of wavelengths. This study examines the usage of PRISMA hyperspectral data to accurately identify and map minerals in Southeast Tamil Nadu’s coastal regions. The raw PRISMA data is refined to reduce atmospheric interference and improve the signal-to-noise ratio (SNR). Advanced spectrum techniques are employed to identify and isolate mineral fingerprints. To increase accuracy in detecting various mineral assemblages, even in complex geological settings, a machine learning (ML) approach incorporating both unsupervised and supervised learning algorithm is utilized. Unsupervised learning methods such as K means clustering, principal component analysis (PCA), and vertex component analysis (VCA) are used for finding similar spectral signatures of mineral. Supervised learning methods such as convolutional neural networks (CNNs), random forests (RFs), logistic regression (LR), support vector machines (SVMs), K nearest neighbors (KNN), decision tree (DT), artificial neural networks (ANNs), linear discriminant analysis (LDA), and Naïve Bayes (NB) are evaluated, and different performance metrics are computed. The results show that CNN model outperforms other ML models interms of overall accuracy (OA) of 78% for mineral classification. Moreover, CNN model could successfully identify and discriminate diverse minerals with subtle spectral differences as well, with testing accuracies of 92% in the Cuddalore region and 94.4% in the Mayiladuthurai region. The validation process was conducted by using the same kind of images. Overall, this study highlights the potential of the use of PRISMA hyperspectral data combined with ML for accurate and detailed mineral identification and mapping in coastal regions.</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

Machine learning-driven mineral identification using PRISMA hyperspectral data along the coastal regions of Southeast Tamil Nadu

  • 期刊论文

  • 发表日期：

    2025年3月14日

查看全文
  • S. Sudharsan,
  • R. Hemalatha,
  • Tejas N. V.,
  • Krisha Aarunee Sivakumar

摘要

Hyperspectral remote sensing has evolved as an effective method for mineral identification and mapping, delivering comprehensive spectral data over a wide range of wavelengths. This study examines the usage of PRISMA hyperspectral data to accurately identify and map minerals in Southeast Tamil Nadu’s coastal regions. The raw PRISMA data is refined to reduce atmospheric interference and improve the signal-to-noise ratio (SNR). Advanced spectrum techniques are employed to identify and isolate mineral fingerprints. To increase accuracy in detecting various mineral assemblages, even in complex geological settings, a machine learning (ML) approach incorporating both unsupervised and supervised learning algorithm is utilized. Unsupervised learning methods such as K means clustering, principal component analysis (PCA), and vertex component analysis (VCA) are used for finding similar spectral signatures of mineral. Supervised learning methods such as convolutional neural networks (CNNs), random forests (RFs), logistic regression (LR), support vector machines (SVMs), K nearest neighbors (KNN), decision tree (DT), artificial neural networks (ANNs), linear discriminant analysis (LDA), and Naïve Bayes (NB) are evaluated, and different performance metrics are computed. The results show that CNN model outperforms other ML models interms of overall accuracy (OA) of 78% for mineral classification. Moreover, CNN model could successfully identify and discriminate diverse minerals with subtle spectral differences as well, with testing accuracies of 92% in the Cuddalore region and 94.4% in the Mayiladuthurai region. The validation process was conducted by using the same kind of images. Overall, this study highlights the potential of the use of PRISMA hyperspectral data combined with ML for accurate and detailed mineral identification and mapping in coastal regions.