An Improved Multi-temporal and Multi-feature Tea Plantation Identification Method Using Sentinel-2 Imagery.

Area of Science:

• Agricultural Remote Sensing
• Geospatial Analysis
• Machine Learning Applications

Background:

• Tea is a vital economic crop requiring efficient monitoring.
• Accurate identification of tea plantations is essential for sustainable practices.
• Existing remote sensing methods need improvement for precision agriculture.

Purpose of the Study:

• To develop and validate a novel method for tea plantation identification.
• To leverage multi-temporal Sentinel-2 imagery and Random Forest algorithms.
• To enhance the accuracy and efficiency of tea crop monitoring.

Main Methods:

• Utilized multi-temporal Sentinel-2 images capturing tea phenological stages.
• Extracted and analyzed spectral bands, spectral derivatives, NDVI, textures, and topographic features.
• Employed Random Forest (RF) algorithm for feature importance assessment and classification.

Main Results:

• The developed RF method achieved high producer's accuracy (96.57%) and user's accuracy (96.02%).
• Feature importance analysis by RF effectively reduced data dimensions and improved classification efficiency.
• Multi-temporal and multi-feature classification significantly enhanced tea plantation recognition accuracy.

Conclusions:

• The combination of multi-temporal Sentinel-2 data and RF algorithm offers a robust solution for tea plantation identification.
• The proposed method provides a reliable tool for sustainable tea management and monitoring.
• This approach demonstrates the potential of advanced remote sensing and machine learning in precision agriculture.