Liu Renhao, Li Qingxu, Zhang Hongzhou, Wang Yinpeng
College of Mechanical and Electrical Engineering, Tarim University, Alar, China.
College of Computer Science, Anhui University of Finance and Economics, Bengbu, China.
J Food Sci. 2025 Aug;90(8):e70477. doi: 10.1111/1750-3841.70477.
Prunes, known for their sweet aroma and smooth texture, are rich in vitamins and minerals, making them both a delicious and nutritious food. Brix is a key indicator of prune quality; however, rapid and nondestructive measurement remains a challenge. About this research, visible/near-infrared (Vis/NIR) spectral data from various prune types were collected and preprocessed using Standard Normal Variate (SNV) to reduce noise and scattering effects. Subsequently, the competitive adaptive reweighted sampling (CARS) algorithm was employed to select feature wavelengths indicative of the brix in prunes. Three predictive models, Partial Least Squares Regression (PLSR), Support Vector Regression (SVR), and Least Squares Support Vector Machine (LSSVM), were developed for prune Brix estimation. Among these, the SVR model, using CARS-selected wavelengths, achieved the highest predictive accuracy, with R (Coefficient of Determination) = 0.8625, RPD (Residual Prediction Deviation) = 2.7229, and RMSE (Root Mean Square Error) = 1.6801. Validation tests further confirmed these results, with an R (Coefficient of Determination) of 0.8608, RPD (Residual Prediction Deviation) of 2.7069, and RMSE (Root Mean Square Error) of 1.69, demonstrating the reliability of the proposed method. This study presents an effective approach for the rapid, nondestructive detection of prune brix levels, providing a valuable technical foundation for future development of fast detection instruments.
李子干以其甜美的香气和顺滑的口感而闻名，富含维生素和矿物质，是一种美味又营养的食物。糖度是李子干品质的关键指标；然而，快速且无损的测量仍是一项挑战。关于这项研究，收集了来自各种李子干类型的可见/近红外（Vis/NIR）光谱数据，并使用标准正态变量变换（SNV）进行预处理，以减少噪声和散射效应。随后，采用竞争性自适应重加权采样（CARS）算法来选择指示李子干糖度的特征波长。开发了三种预测模型用于李子干糖度估计，分别是偏最小二乘回归（PLSR）、支持向量回归（SVR）和最小二乘支持向量机（LSSVM）。其中，使用CARS选择波长的SVR模型实现了最高的预测精度，决定系数R = 0.8625，残差预测偏差RPD = 2.7229，均方根误差RMSE = 1.6801。验证测试进一步证实了这些结果，决定系数R为0.8608，残差预测偏差RPD为2.7069，均方根误差RMSE为1.69，证明了所提方法的可靠性。本研究提出了一种快速、无损检测李子干糖度水平的有效方法，为未来快速检测仪器的发展提供了有价值的技术基础。
