Fourier transform infrared spectrum pre-processing technique selection for detecting PYLCV-infected chilli plants

Area of Science:

• Agricultural Science
• Spectroscopy
• Plant Pathology

Background:

• Fourier transform infrared (FTIR) spectroscopy is valuable for analyzing plant health.
• Effective pre-processing is essential to reduce noise and improve accuracy in spectral analysis.
• Detecting plant diseases like pepper yellow leaf curl virus (PYLCV) in chilli plants requires robust analytical methods.

Purpose of the Study:

• To optimize pre-processing techniques for FTIR spectroscopy to detect PYLCV-infected chilli plants.
• To evaluate the impact of different pre-processing methods on classification accuracy.
• To identify the most effective pre-processing strategy for simplified and accurate disease detection.

Main Methods:

• Applied various pre-processing techniques: baseline correction, normalization (SNV, vector, min-max), and de-noising (Savitzky-Golay (SG) smoothing, 1st/2nd derivatives).
• Utilized discrete wavelet transform (DWT) for dimension reduction on spectral data (mid-IR and biofingerprint regions).
• Employed classification algorithms: multilayer perceptron neural network, support vector machine, and linear discriminant analysis.

Main Results:

• The Savitzky-Golay (SG) 1st derivative method, applied to both spectral ranges, achieved 100% classification accuracy.
• Principal component analysis (PCA) clustering supported the high accuracy of the selected pre-processing method.
• Optimized pre-processing simplified the classification process and increased success rates.

Conclusions:

• The correct selection and optimization of pre-processing techniques are critical for enhancing FTIR spectroscopy-based plant disease detection.
• The SG 1st derivative method offers a highly effective and accurate approach for identifying PYLCV infection in chilli plants.
• This study demonstrates the potential of optimized FTIR spectroscopy for efficient and accurate agricultural disease diagnostics.