UAV and Machine Learning Based Refinement of a Satellite-Driven Vegetation Index for Precision Agriculture

Area of Science:

• Agricultural Science
• Remote Sensing
• Computer Science

Background:

• Precision agriculture requires accurate crop status data for site-specific management.
• Satellite imagery is valuable for crop monitoring but low-resolution data can be inaccurate for row crops.
• Unmanned Aerial Vehicle (UAV) multispectral sensors offer high-resolution data but are often costly.

Purpose of the Study:

• To develop a cost-effective framework for refining satellite imagery using deep learning.
• To enhance the accuracy of crop status assessment in precision agriculture.
• To create detailed vineyard vigor maps for growers.

Main Methods:

• A novel deep learning framework was developed to refine satellite imagery.
• The framework utilizes high-resolution data from Unmanned Aerial Vehicle (UAV) multispectral sensors.
• A convolutional neural network was trained on a single UAV dataset.
• A case study in a Northern Italian vineyard was used for validation.

Main Results:

• Refined satellite-derived Normalized Difference Vegetation Index (NDVI) maps showed improved accuracy in describing crop status compared to raw data.
• Correlation analysis and ANOVA confirmed the enhanced accuracy of the refined maps.
• A K-means classifier successfully generated 3-class vineyard vigor maps from the refined NDVI data.

Conclusions:

• The proposed deep learning framework effectively refines satellite imagery for precision agriculture.
• The method is simple, cost-effective, and improves crop status assessment accuracy.
• The generated vigor maps are valuable tools for vineyard management and sustainable practices.