[Prediction the Soluble Solid Content in Sugarcanes by Using Near Infrared Hyperspectral Imaging System]

Area of Science:

• Agricultural Science
• Spectroscopy
• Data Science

Context:

• Soluble solid content (SSC) is a crucial quality parameter in sugarcane.
• Accurate and non-destructive methods for SSC prediction are highly desirable in the agricultural industry.
• Near-infrared (NIR) hyperspectral imaging offers potential for rapid, in-situ quality assessment.

Purpose:

• To investigate the feasibility of predicting sugarcane SSC using NIR hyperspectral imaging.
• To compare the performance of different chemometric models (PLSR, PCR, LS-SVM) for SSC prediction.
• To evaluate the effectiveness of wavelength selection algorithms (SPA, iPLS, UVE) in improving prediction accuracy.

Summary:

• Two hundred and forty sugarcane stalks from three varieties were analyzed using NIR hyperspectral imaging.
• Prediction models were developed using Partial Least Squares Regression (PLSR), Principal Components Regression (PCR), and Least Squares Support Vector Machines (LS-SVM).
• The PLSR model based on spectral features demonstrated robust performance, with R² values of 0.879 (calibration) and 0.843 (prediction).
• Wavelength selection using the Uninformative Variables Elimination (UVE) algorithm identified 105 effective wavelengths, leading to a PLSR model with R² of 0.860 (calibration) and 0.813 (prediction).

Impact:

• This study validates NIR hyperspectral imaging as a viable technique for non-destructive SSC prediction in sugarcane.
• The findings provide a foundation for developing advanced quality control systems in sugarcane processing.
• Optimized models using selected wavelengths offer potential for cost-effective and efficient quality assessment.