Research on Classification of Open-Pit Mineral Exploiting Information Based on OOB RFE Feature Optimization

Area of Science:

• Earth and Space Sciences
• Geosciences
• Remote Sensing

Background:

• Accurate extraction of mineral exploiting information is crucial for regional mineral activities, management, and environmental protection.
• Multi-source remote sensing data is increasingly used for land surface classification, but optimal feature combination selection remains a challenge.

Purpose of the Study:

• To creatively combine Out of Bag data with Recursive Feature Elimination (OOB RFE) to optimize feature combinations for non-metallic building material mineral exploiting information.
• To assess the effectiveness of the OOB RFE method combined with random forest (RF) for accurate classification of mineral exploiting information.

Main Methods:

• Acquired and integrated Ziyuan-1-02D (ZY-1-02D) hyperspectral, Landsat-8 multispectral, and Sentinel-1 Synthetic Aperture Radar (SAR) imagery.
• Extracted spectrum, heat, polarization, and texture features.
• Employed OOB RFE for feature optimization and random forest (RF) for classification.

Main Results:

• The OOB RFE method combined with RF achieved the highest overall accuracy of 93.64% (kappa coefficient of 0.926).
• OOB RFE precisely filtered feature combinations, leading to optimal classification results compared to RFE alone.
• RF proved effective in classifying mineral exploiting information under the optimized feature scheme.

Conclusions:

• The proposed OOB RFE feature optimization method and optimal feature combination provide effective technical support for mineral exploiting information extraction.
• This approach offers a theoretical reference for mineral exploiting information classification in other regions.
• The study highlights the importance of feature selection for accurate remote sensing-based land surface classification.