Scholarlink

<p>To develop accurate and efficient machine learning models for land cover classification using satellite imagery. This research investigates the application of deep convolutional neural networks (DCNNs) for accurate and efficient land cover classification using multispectral Landsat satellite imagery. The study explores the performance of three DCNN architectures, AlexNet, DenseNet, and VGG-16, on a comprehensive Landsat dataset, comparing their efficacy against traditional machine learning methods, including a Support Vector Machine (SVM) and Random Forest. Data preprocessing, feature extraction, and hyperparameter tuning were implemented to model performance. Our findings demonstrate that DCNN-based models, particularly DenseNet and VGG-16, significantly outperform traditional techniques in terms of accuracy, precision, and recall. Specifically, DenseNet exhibits the most substantial improvements, surpassing the SVM baseline. These results highlight the potential of deep learning for advancing land cover classification from satellite data, enabling more rapid and precise identification of land use categories, such as woodlands, wetlands, and built-up areas. The study also suggests the feasibility of adaptive learning strategies, potentially incorporating reinforcement learning and geospatial information, to further enhance classification efficiency.</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

Satellite imagery for land cover classification using machine learning techniques

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Ramalingam Sugumar,
  • D. Suganya

摘要

To develop accurate and efficient machine learning models for land cover classification using satellite imagery. This research investigates the application of deep convolutional neural networks (DCNNs) for accurate and efficient land cover classification using multispectral Landsat satellite imagery. The study explores the performance of three DCNN architectures, AlexNet, DenseNet, and VGG-16, on a comprehensive Landsat dataset, comparing their efficacy against traditional machine learning methods, including a Support Vector Machine (SVM) and Random Forest. Data preprocessing, feature extraction, and hyperparameter tuning were implemented to model performance. Our findings demonstrate that DCNN-based models, particularly DenseNet and VGG-16, significantly outperform traditional techniques in terms of accuracy, precision, and recall. Specifically, DenseNet exhibits the most substantial improvements, surpassing the SVM baseline. These results highlight the potential of deep learning for advancing land cover classification from satellite data, enabling more rapid and precise identification of land use categories, such as woodlands, wetlands, and built-up areas. The study also suggests the feasibility of adaptive learning strategies, potentially incorporating reinforcement learning and geospatial information, to further enhance classification efficiency.