Combining Hyperspectral Imaging with Ensemble Learning for Estimating Rapeseed Chlorophyll Content Under Different...

Area of Science:

• Agricultural Science
• Plant Physiology
• Remote Sensing

Background:

• Chlorophyll content is vital for photosynthetic capacity and crop health.
• Soil-Plant Analysis Development (SPAD) meters offer rapid, non-destructive chlorophyll estimation.
• Hyperspectral imaging (HSI) captures detailed spectral data for advanced crop analysis.

Purpose of the Study:

• To assess the efficacy of HSI integrated with ensemble learning (EL) for estimating rapeseed seedling SPAD values.
• To evaluate model performance under varying waterlogging durations and across different rapeseed cultivars.
• To establish a robust method for early stress detection in crops.

Main Methods:

• Collected hyperspectral images and SPAD values from six rapeseed cultivars subjected to 0, 2, 4, and 6 days of waterlogging.
• Utilized mutual information for selecting the top 30 spectral and vegetation index features.
• Developed an EL model using five first-layer learners (PLS, SVM, RF, Ridge, Elastic Net) and a second-layer multiple linear regression.

Main Results:

• The EL model demonstrated superior stability and prediction accuracy over single models across diverse datasets.
• Model accuracy improved with increased waterlogging duration, reaching an R² of 0.79 and RMSE of 3.27 at 6 days.
• The developed model showed strong predictive capability for estimating SPAD values.

Conclusions:

• Combining ensemble learning with hyperspectral imaging provides a stable and accurate method for estimating chlorophyll content.
• This integrated approach offers a powerful tool for early stress monitoring in agricultural settings.
• The study highlights the potential of advanced spectral analysis for high-throughput phenotyping and crop management.