Sugarcane nitrogen nutrition estimation with digital images and machine learning methods.

Area of Science:

• Agricultural Science
• Plant Physiology
• Digital Image Processing

Background:

• Crop nutrient status, particularly nitrogen (N), is crucial for yield and can be inferred from leaf characteristics.
• Optimizing nitrogen fertilizer management is essential for sustainable agriculture and environmental protection.

Purpose of the Study:

• To develop and validate machine learning models for estimating sugarcane leaf nitrogen content using digital image analysis.
• To compare the performance of models based on color features, texture features, and integrated features.

Main Methods:

• Collected sugarcane leaf images at tillering and elongation stages using a digital camera.
• Extracted color features (CF) and texture features (TF) using digital image processing.
• Applied principal component analysis (PCA) for feature dimensionality reduction.
• Developed N content estimation models using multiple linear regression (MLR), random forest (RF), and stacking fusion model (SFM).

Main Results:

• Models integrating color and texture principal component features (C-T-PCA) outperformed single-feature models.
• The SFM achieved the highest accuracy, with R² values of 0.9264 (tillering) and 0.9111 (elongation).
• SFM demonstrated significant improvements in accuracy compared to other models.

Conclusions:

• The SFM framework based on C-T-PCA effectively enhances model performance, anti-interference, and generalization capabilities.
• Digital image processing combined with machine learning provides a rapid and non-destructive method for assessing crop nitrogen nutrition.
• This approach facilitates optimized nitrogen fertilizer management in sugarcane.