[Research on spectra recognition method for cabbages and weeds based on PCA and SIMCA]

Area of Science:

• Agricultural Science
• Spectroscopy
• Machine Learning

Background:

• Effective weed identification is vital for crop yield and management.
• Spectral reflectance analysis offers a non-destructive method for plant discrimination.
• Existing methods require optimization for accuracy and efficiency.

Purpose of the Study:

• To develop an accurate and efficient method for distinguishing weeds from crops using spectral data.
• To optimize spectral data preprocessing techniques for enhanced classification.
• To identify key spectral features for robust plant classification.

Main Methods:

• Applied Savitzky-Golay (SG) convolutional derivation and Multiplicative Scattering Correction (MSC) for spectral data preprocessing.
• Utilized Principal Component Analysis (PCA) for clustering analysis and feature wavelength extraction.
• Employed Soft Independent Modeling of Class Analogy (SIMCA) for plant classification using selected feature wavelengths.

Main Results:

• Optimal preprocessing involved MSC combined with SG derivation (1st order, 3rd degree polynomial, 51 smoothing points).
• 23 sensitive feature wavelengths were identified using PCA.
• SIMCA achieved high classification rates: 98.6% for the modeling set and 100% for the prediction set.

Conclusions:

• The combined MSC and SG preprocessing method significantly improves spectral data quality for classification.
• PCA effectively identifies crucial wavelengths for distinguishing between plant species.
• SIMCA, utilizing selected feature wavelengths, provides a highly accurate method for weed identification in agricultural settings.