Study on early rice blast diagnosis based on unpre-processed Raman spectral data

Area of Science:

• Agricultural Science
• Spectroscopy
• Biotechnology

Background:

• Traditional rice blast diagnosis relies on visual inspection, lacking early detection capabilities.
• Early identification of rice blast is crucial for effective disease management and yield preservation.
• Complex biological samples present challenges in Raman spectroscopy due to overlapping peaks, fluorescence, and noise.

Purpose of the Study:

• To develop a robust method for early rice blast diagnosis using Raman spectroscopy.
• To establish a classification model directly from raw spectral data, bypassing conventional pre-processing steps.
• To evaluate the diagnostic accuracy of the proposed method compared to existing pre-processing techniques.

Main Methods:

• Raw Raman spectral data were decomposed using Empirical Mode Decomposition (EMD) into Intrinsic Mode Functions (IMFs).
• Signal components were extracted by filtering IMFs based on self-correlation and zero-crossing analysis.
• Characteristic variables (β-carotene, chlorophyll, chitin) were screened using Successive Projections Algorithm (SPA).
• A Partial Least Squares (PLS) regression model was built for rice blast classification.

Main Results:

• The proposed method achieved a test classification accuracy of 94.12% for rice blast detection.
• This accuracy surpasses models that utilized pre-processing methods like Moving Average Smoothing, Savitzky Golay Smoothing, Gaussian Filter Smoothing, and various wavelet-based techniques.
• The direct modeling of raw data with EMD and SPA proved effective in handling spectral complexity.

Conclusions:

• Empirical Mode Decomposition combined with Successive Projections Algorithm offers a superior approach for analyzing complex Raman spectra in biological samples.
• This method enables accurate early detection of rice blast without the need for extensive spectral pre-processing.
• The developed Raman spectroscopy technique holds significant potential for practical application in agricultural disease diagnostics.