Prediction of soluble solids content and anthocyanin content in blood oranges based on hyperspectral reflectance and transmittance imaging technologies

文献类型： 外文期刊

第一作者： Liu, Xuefeng

作者： Liu, Xuefeng;Xiang, Pingwei;Qin, Junlian;Ma, Xiaoli;Yuan, Xiangcheng;Hu, Deyu;Tian, Xi;Tian, Xi

作者机构：

关键词： hyperspectral imaging; blood oranges; soluble solids content; anthocyanin content

期刊名称：INTERNATIONAL JOURNAL OF FOOD ENGINEERING （ 影响因子：1.4； 五年影响因子：2.1 ）

ISSN： 2194-5764

年卷期： 2025 年 21 卷 8 期

页码：

收录情况： SCI

摘要： Anthocyanins and soluble solids content (SSC) serve as key factors for evaluating blood orange quality. Currently, reliable non-destructive measurement methods are lacking in production. In this study, hyperspectral diffuse reflectance and transmittance imaging (400 nm-1,000 nm) technologies were utilized to predict SSC and anthocyanin content in blood oranges. Three methods including standard normal variate (SNV) correction, moving average smoothing (MAS), and first derivative (Deriv1) were employed for preprocessing spectra. Additionally, bootstrapping soft shrinkage (BOSS), successive projections algorithm (SPA), and competitive adaptive reweighted sampling (CARS) were used to select effective wavelengths. Finally, partial least squares regression (PLSR) models were developed for predicting anthocyanin content and SSC in blood oranges. The results showed that the hyperspectral transmittance imaging mode exhibited higher accuracy in predicting SSC and anthocyanin content in blood oranges when compared to the diffuse reflectance mode. Among the tested conditions, preprocessing the original spectra with SNV and establishing a PLSR model utilizing full-wavelength spectrum yielded the highest prediction accuracy for SSC, where Rpre was 0.927, RMSEP was 0.418 degrees Brix, and RPD was 2.621. On the other hand, preprocessing the original spectra with SNV and establishing a PLSR model with SPA-selected effective wavelengths exhibited optimal performance in predicting anthocyanin content, where Rpre was 0.872, RMSEP was 1.702 mg/100 mL, and RPD was 1.918. Additionally, the spatial distributions of SSC and anthocyanin content in blood oranges were visualized using the optimal models. The findings demonstrate that hyperspectral imaging combined with effective spectral preprocessing and wavelength extraction algorithms can achieve non-destructive quality prediction of blood oranges.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序