The Inversion of SPAD Value in Pear Tree Leaves by Integrating Unmanned Aerial Vehicle Spectral Information and...

Area of Science:

• Precision Agriculture
• Remote Sensing
• Plant Physiology

Background:

• Chlorophyll content is vital for pear tree health and fruit yield.
• Accurate estimation of leaf chlorophyll (SPAD) values is essential for effective orchard management.
• Unmanned Aerial Vehicle (UAV) multispectral imagery offers a non-destructive method for assessing plant health.

Purpose of the Study:

• To integrate UAV-derived multispectral vegetation indices and textural features for estimating pear leaf SPAD values.
• To compare the performance of different machine learning models (XGBoost, RF, BPNN, OIA) in SPAD value inversion.
• To determine the optimal combination of features and models for accurate SPAD estimation.

Main Methods:

• Extraction of vegetation indices and textural features from UAV multispectral images during pear fruit expansion.
• Analysis of correlations between extracted features and ground-measured SPAD values.
• Development and comparison of four machine learning models (XGBoost, RF, BPNN, OIA) using various feature combinations.

Main Results:

• Both vegetation indices and textural features showed significant correlations with SPAD values.
• Combining vegetation indices and textural features substantially improved SPAD estimation accuracy over single feature types.
• The Optimized Integration Algorithm (OIA) model, using combined features, achieved the highest accuracy (R²=0.931 training, R²=0.877 validation).

Conclusions:

• Integrated multispectral vegetation indices and textural features enhance SPAD value estimation accuracy.
• Machine learning models, particularly the OIA, effectively invert SPAD values using combined features.
• This approach supports refined, data-driven management strategies for pear orchards.