Application of Hyperspectral Technology with Machine Learning for Brix Detection of Pastry Pears
Hongkun Ouyang1, Lingling Tang1, Jinglong Ma1
1College of Mechanical and Electrical Engineering, Sichuan Agriculture University, Ya'an 625000, China.
Plants (Basel, Switzerland)
|April 27, 2024
Summary
This study developed a genetically tuned support vector machine model using hyperspectral technology to accurately estimate crisp pear sugar content. The SG-CARS-GASVR model achieved a high correlation coefficient, improving fruit quality assessment.
Area of Science:
- Agricultural Science
- Analytical Chemistry
- Machine Learning
Background:
- Sugar content is crucial for crisp pear quality, impacting identification and market predictions.
- Accurate and efficient methods are needed for assessing pear sugar levels.
Purpose of the Study:
- To develop a reliable hyperspectral method for estimating crisp pear sugar content.
- To optimize a support vector machine (SVM) model for improved accuracy.
Main Methods:
- Evaluated spectral data against actual sugar content in crisp pears.
- Applied preprocessing techniques: standard normal variable transformation (SNV), multivariate scattering correction (MSC), and convolution smoothing (SG).
- Utilized genetic algorithms (GA), competitive adaptive reweighting (CARS), and continuous projection algorithm (SPA) for wavelength selection and model optimization.
Main Results:
- The SVM model preprocessed with convolution smoothing (SG) showed improved accuracy.
- The genetically tuned SVM (GASVR) model with SG preprocessing enhanced prediction accuracy.
- The SG-CARS-GASVR model achieved the highest correlation coefficient (0.8992) and reduced prediction error.
Conclusions:
- The developed SG-CARS-GASVR model offers a reliable and efficient method for detecting crisp pear sugar content using hyperspectral technology.
- This approach enhances the accuracy and efficiency of crisp pear quality assessment.
