Hyperspectral and Fluorescence Imaging Approaches for Nondestructive Detection of Rice Chlorophyll

Area of Science:

• Agricultural Science
• Remote Sensing
• Spectroscopy

Background:

• Chlorophyll content estimation is vital for rice crop management, impacting fertilization, yield, and quality.
• Non-destructive methods for assessing chlorophyll are crucial for optimizing agricultural practices and sustainability.

Purpose of the Study:

• To develop an accurate, non-destructive method for estimating chlorophyll content in rice leaves using spectral analysis.
• To evaluate the effectiveness of various feature extraction and machine learning algorithms for this purpose.

Main Methods:

• Collected ninety experimental spectral datasets from rice leaf samples.
• Applied feature extraction algorithms to compress spectral bands and reveal correlations.
• Constructed and evaluated machine learning models (including CNN+LSTM) using hyperspectral and fluorescence spectroscopy data.

Main Results:

• The Iteratively Variable Subset Optimization-Interval Variable Iterative Space Shrinkage Approach (IVSO-IVISSA) combined with fluorescence data and CNN+LSTM achieved the best prediction performance.
• Achieved RMSE-Train of 0.26, RMSE-Test of 0.29, and RPD of 2.64 with the optimal method.
• Demonstrated the efficacy of spectral analysis combined with feature extraction and machine learning for chlorophyll estimation.

Conclusions:

• Hyperspectral and fluorescence spectroscopy, coupled with advanced analytical techniques, provide a novel approach for rapid, non-destructive crop health monitoring.
• This methodology is critical for advancing smart and precision agriculture, enabling optimized resource management.