Classification of Nitrogen-Efficient Wheat Varieties Based on UAV Hyperspectral Remote Sensing.

Area of Science:

• Agricultural Science
• Remote Sensing
• Machine Learning

Background:

• Traditional wheat variety classification is inefficient and labor-intensive.
• Nitrogen use efficiency is critical for sustainable agriculture and crop yield.
• Unmanned Aerial Vehicle (UAV) hyperspectral remote sensing offers a promising non-destructive approach for crop analysis.

Purpose of the Study:

• To develop an efficient and accurate method for classifying nitrogen-efficient wheat varieties.
• To leverage UAV hyperspectral remote sensing data for improved wheat breeding and nitrogen management.
• To overcome limitations of traditional classification methods in terms of time, cost, and labor.

Main Methods:

• Utilized t-SNE dimensionality reduction and hierarchical clustering to analyze agronomic indicators and classify 12 wheat varieties.
• Employed Least Absolute Shrinkage and Selection Operator-Competitive Adaptive Reweighted Sampling (Lasso-CARS) for hyperspectral feature band selection.
• Developed a Support Vector Machine-Extreme Gradient Boosting (SVM-XGBoost) model integrating SVM outputs with XGBoost for classification.

Main Results:

• The SVM-XGBoost model achieved high classification accuracies under varying nitrogen stress conditions (74% low, 83% high, 70% no nitrogen).
• The method effectively extracted relevant hyperspectral bands, mitigating data collinearity and noise.
• The model successfully classified wheat varieties and informed nitrogen fertilization strategies.

Conclusions:

• The proposed UAV hyperspectral remote sensing-based SVM-XGBoost method provides an efficient and accurate approach for nitrogen-efficient wheat variety classification.
• This technology supports precision breeding efforts and optimizes nitrogen management in wheat cultivation.
• The study offers a foundation for accelerating the development of nitrogen-efficient crop varieties.