Vegetation Classification in a Mountain–Plain Transition Zone in the Sichuan Basin, China DOI Creative Commons
Wenqian Bai, Zhengwei He, Yan Tan

et al.

Land, Journal Year: 2025, Volume and Issue: 14(1), P. 184 - 184

Published: Jan. 17, 2025

Developing an effective vegetation classification method for mountain–plain transition zones is critical understanding ecological patterns, evaluating ecosystem services, and guiding conservation efforts. Existing methods perform well in mountainous plain areas but lack verification zones. This study utilized terrain data Sentinel-1 Sentinel-2 imagery to extract topographic, spectral, texture, SAR features as the index. By combining feature sets applying elimination algorithms, performance of one-dimensional convolutional neural networks (1D-CNNs), Random Forest (RF), Multilayer Perceptron (MLP) was evaluated determine optimal combinations methods. The results show following: (1) multi-feature combinations, especially spectral topographic features, significantly improved accuracy; (2) Recursive Feature Elimination based on (RF-RFE) outperformed ReliefF selection, identifying more representative features; (3) all three algorithms performed well, with consistent spatial results. MLP algorithm achieved best overall accuracy (OA: 81.65%, Kappa: 77.75%), demonstrating robustness lower dependence quantity. presents efficient robust workflow, verifies its applicability zones, provides valuable insights small-region under similar conditions globally.

Language: Английский

Vegetation Classification in a Mountain–Plain Transition Zone in the Sichuan Basin, China DOI Creative Commons
Wenqian Bai, Zhengwei He, Yan Tan

et al.

Land, Journal Year: 2025, Volume and Issue: 14(1), P. 184 - 184

Published: Jan. 17, 2025

Developing an effective vegetation classification method for mountain–plain transition zones is critical understanding ecological patterns, evaluating ecosystem services, and guiding conservation efforts. Existing methods perform well in mountainous plain areas but lack verification zones. This study utilized terrain data Sentinel-1 Sentinel-2 imagery to extract topographic, spectral, texture, SAR features as the index. By combining feature sets applying elimination algorithms, performance of one-dimensional convolutional neural networks (1D-CNNs), Random Forest (RF), Multilayer Perceptron (MLP) was evaluated determine optimal combinations methods. The results show following: (1) multi-feature combinations, especially spectral topographic features, significantly improved accuracy; (2) Recursive Feature Elimination based on (RF-RFE) outperformed ReliefF selection, identifying more representative features; (3) all three algorithms performed well, with consistent spatial results. MLP algorithm achieved best overall accuracy (OA: 81.65%, Kappa: 77.75%), demonstrating robustness lower dependence quantity. presents efficient robust workflow, verifies its applicability zones, provides valuable insights small-region under similar conditions globally.

Language: Английский

Citations

0