Scholarlink

<p>In this study, Visible-Near-Infrared (VIS-NIR) spectroscopy was employed for the rapid and non-destructive detection of fig ripeness.The partial least squares regression (PLSR) method was used as the primary modeling technique, while the random forest (RF) method, Gaussian process regression (GPR) method, and extreme gradient boosting (XGBoost) method served as comparative methods. Meanwhile, the effects of eight spectral preprocessing methods and three wavelength selection algorithms on the prediction of fig color, soluble solid content, and hardness were evaluated, and the performances of PLSR, RF, GPR, and XGBoost were compared. The results show that the PLSR method is significantly better than the other three method. Standardization was identified as the optimal preprocessing method, and the competitive adaptive reweighted sampling (CARS) algorithm was selected as the optimal wavelength selection method. The CARS-PLSR achieved the best prediction results ( R²_train = 0.9816, RMSE_train = 0.1443, R²_test = 0.9637, RMSE_test = 0.2284, RPD = 4.8518, RRMSE = 8.54%). Independent validation was conducted using fig samples harvested in 2024 (<i>n</i> = 30), further confirming the superiority of the standardized competitive adaptive reweighted sampling partial least squares model (R² = 0.8499, RMSE = 0.7965).This study is the first to establish a fig maturity model based on the a* value and conduct cross-year validation, providing a feasible technical solution for commercial fig detection.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Nondestructive VIS-NIR prediction of fig maturity primarily using peel color

  • 期刊论文

  • 发表日期：

    2026年4月22日

查看全文
  • Mingyue Pang,
  • Yanling Han,
  • Yangguang Wang,
  • Rui Sun

摘要

In this study, Visible-Near-Infrared (VIS-NIR) spectroscopy was employed for the rapid and non-destructive detection of fig ripeness.The partial least squares regression (PLSR) method was used as the primary modeling technique, while the random forest (RF) method, Gaussian process regression (GPR) method, and extreme gradient boosting (XGBoost) method served as comparative methods. Meanwhile, the effects of eight spectral preprocessing methods and three wavelength selection algorithms on the prediction of fig color, soluble solid content, and hardness were evaluated, and the performances of PLSR, RF, GPR, and XGBoost were compared. The results show that the PLSR method is significantly better than the other three method. Standardization was identified as the optimal preprocessing method, and the competitive adaptive reweighted sampling (CARS) algorithm was selected as the optimal wavelength selection method. The CARS-PLSR achieved the best prediction results ( R²_train = 0.9816, RMSE_train = 0.1443, R²_test = 0.9637, RMSE_test = 0.2284, RPD = 4.8518, RRMSE = 8.54%). Independent validation was conducted using fig samples harvested in 2024 (n = 30), further confirming the superiority of the standardized competitive adaptive reweighted sampling partial least squares model (R² = 0.8499, RMSE = 0.7965).This study is the first to establish a fig maturity model based on the a* value and conduct cross-year validation, providing a feasible technical solution for commercial fig detection.