数字农科院2.0

Quantitative estimation of blueberry SSC using fractional order derivative coupled optimized spectral indices

文献类型： 外文期刊

作者： Anhong Tian;Han Zhang;Chengbiao Fu;Zhiyong Cao;Denghua Li

作者机构：

关键词： Blueberry SSC;BPNN;Fractional order derivative;Optimized spectral indices;Vis-NIR spectroscopy

期刊名称： Measurement: Journal of the International Measurement Confederation

ISSN： 0263-2241

年卷期： 2025 年 257 卷

页码：

收录情况： SCIE（2025版） ; ; EI（2025版）

摘要： The soluble solids content (SSC) is a pivotal metric for assessing the internal quality of blueberries. The development of rapid and non-destructive SSC prediction techniques has paramount importance for the grading of blueberries post-harvest. In this study, we propose a feature selection strategy that integrates the fractional-order derivative (FOD) and two-dimensional spectral indices and systematically compare the applicability of different algorithms in blueberry SSC prediction. A total of 130 sets of blueberry samples at different maturity stages (ripe, immature, and semi-ripe) from a plantation area in Qujing City, Yunnan Province, were used for the study. The visible and near-infrared (Vis-NIR) spectroscopy of blueberries was collected and preprocessed with a forward-oriented derivative (FOD) of 0 to 2 orders (step = 0.2) to construct four two-dimensional (2D)-optimized spectral indices. These indices were used to screen characteristic bands and establish prediction models for a partial least squares regression (PLSR) and a back propagation neural network (BPNN), respectively. The results of the study indicate that: (1) The FOD-coupled spectral indices have been shown to enhance the feature extraction ability. (2) The following characteristic bands have been identified as SSC-sensitive: 778 nm, 727 nm, 679 nm, and 546 nm. (3) The performance of the validation in models shows an overall trend of increasing and then decreasing with the increase of FOD. Finally, the PLSR reaches the optimum at the 0.8-order derivative (R2 = 0.764, RPD = 2.058) and the BPNN performed optimally in the order of the 0.6-order derivative (R2 = 0.852, RPD = 2.599). These results are both significantly improved compared with the traditional integer-order derivatives. Furthermore, the BPNN has an overall better prediction accuracy than the PLSR by virtue of its nonlinear mapping capability. The study's findings substantiate the efficacy of a combined method based on the FOD-spectral indices in enhancing the prediction accuracy of blueberry SSC. This method provides a theoretical foundation and an optimization approach for expeditious and nondestructive prediction of blueberry SSC.

分类号：

  • 相关文献

[1]Identification of pummelo cultivars by using Vis/NIR spectra and pattern recognition methods. Li, Xun-lan,Li, Xun-lan,Li, Xun-lan,Yi, Shi-lai,He, Shao-lan,Lv, Qiang,Xie, Rang-jin,Zheng, Yong-qiang,Deng, Lie.

[2]Modeling the Mechanical Properties of Root–Substrate Interaction with a Transplanter Using Artificial Neural Networks. Zhiwei Tian,Ang Gao,Wei Ma,Huanyu Jiang,Dongping Cao,Weizi Wang,Jianping Qian,Lijia Xu. 2024

作者其他论文 更多>>