Automatic classification of fine-scale mountain vegetation based on mountain altitudinal belt

Area of Science:

• Geoscience
• Mountain Ecology
• Remote Sensing

Background:

• Fine-scale vegetation mapping is crucial for geoscience and mountain ecology.
• Improved remote sensing resolution enables detailed vegetation mapping.
• Automatic sample collection remains a bottleneck for vegetation classification.

Purpose of the Study:

• To develop an automated method for fine-scale mountain vegetation classification and sample selection.
• To address the challenge of real-time and rapid sample collection.
• To improve the efficiency and accuracy of vegetation mapping in large-scale areas.

Main Methods:

• Stratified sampling based on Mountain Altitudinal Belt (MAB) distribution to divide the study area.
• Automatic selection and correction of samples using iterative clustering within each belt.
• Application of the robust Random Forest (RF) classifier for automatic vegetation classification.

Main Results:

• Achieved an average sample accuracy of 0.933 for nine vegetation formations.
• Obtained a total classification accuracy of 92.2% with a kappa coefficient of 0.910.
• Demonstrated the method's ability to automatically select high-quality samples and produce accurate vegetation maps.

Conclusions:

• The proposed method effectively automates sample selection and fine-scale mountain vegetation classification.
• This approach significantly enhances efficiency compared to traditional vegetation mapping techniques.
• The method holds great significance for large-scale, fine-scale mountain vegetation mapping.