Scholarlink

<p>Rapid urbanization and environmental pressures cause heat stress, loss of open spaces, expansion of impervious surfaces, and flood risk, highlighting the need for precise spatial data for resilient planning and sustainable urban development. Hyperspectral remote sensing integrated with machine learning (ML) offers new opportunities to understand complex urban land use and land cover (LULC) patterns. This study evaluates the potential of PRISMA hyperspectral imagery for characterizing LULC in a diverse urban environment. We employed six machine learning classifiers: Random Forest (RF), Gradient Tree Boosting (GTB), Artificial Neural Network (ANN), Support Vector Machine (SVM), Classification and Regression Tree (CART), and K-nearest Neighbors (KNN) to map different land cover types. The algorithms used 1712 sample points, with a split ratio of 70% for training and 30% for validation, along with extensive hyperparameter tuning. The performance of these algorithms were evaluated using Overall Accuracy (OA), Kappa index, and performance overall (POA), based on metrics such as recall, F1-score, precision, accuracy (ACC), and Matthews correlation coefficient (MCC). The results demonstrate that PRISMA hyperspectral imagery can accurately identify complex land cover types across urban landscapes, with the SVM classifier outperforming others (OA = 93.16%, POA = 2.8087), compared to ANN, RF, and GTB classifiers. Urban classes were discriminated with high accuracy, demonstrating the potential of PRISMA to recognize heterogeneous built-up conditions. The findings of this study also demonstrate how PRISMA can provide the geospatial information for policy, enabling urban monitoring and supporting climate-sensitive, resilience-focused urban planning.</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

Urban land cover characterization using PRISMA hyperspectral imagery and machine learning algorithms

  • 期刊论文

  • 发表日期：

    2026年6月2日

查看全文
  • Zohaib,
  • Sawaid Abbas

摘要

Rapid urbanization and environmental pressures cause heat stress, loss of open spaces, expansion of impervious surfaces, and flood risk, highlighting the need for precise spatial data for resilient planning and sustainable urban development. Hyperspectral remote sensing integrated with machine learning (ML) offers new opportunities to understand complex urban land use and land cover (LULC) patterns. This study evaluates the potential of PRISMA hyperspectral imagery for characterizing LULC in a diverse urban environment. We employed six machine learning classifiers: Random Forest (RF), Gradient Tree Boosting (GTB), Artificial Neural Network (ANN), Support Vector Machine (SVM), Classification and Regression Tree (CART), and K-nearest Neighbors (KNN) to map different land cover types. The algorithms used 1712 sample points, with a split ratio of 70% for training and 30% for validation, along with extensive hyperparameter tuning. The performance of these algorithms were evaluated using Overall Accuracy (OA), Kappa index, and performance overall (POA), based on metrics such as recall, F1-score, precision, accuracy (ACC), and Matthews correlation coefficient (MCC). The results demonstrate that PRISMA hyperspectral imagery can accurately identify complex land cover types across urban landscapes, with the SVM classifier outperforming others (OA = 93.16%, POA = 2.8087), compared to ANN, RF, and GTB classifiers. Urban classes were discriminated with high accuracy, demonstrating the potential of PRISMA to recognize heterogeneous built-up conditions. The findings of this study also demonstrate how PRISMA can provide the geospatial information for policy, enabling urban monitoring and supporting climate-sensitive, resilience-focused urban planning.