您好，欢迎访问浙江省农业科学院 机构知识库！
浙江省农业科学院机构知识库

全部

  • 全部
  • 标题
  • 学者
  • 机构
  • 关键词

Nondestructive detection of multiple dried squid qualities by hyperspectral imaging combined with 1D-KAN-CNN

文献类型： 外文期刊

作者： Hu, Jun 1 ; Jiang, Yuanhao 2 ; Sun, Dawei 1 ; Zhou, Hongkui 1 ; Lou, Weidong 1 ; Zhang, Jin 1 ; Zhou, Chengquan 1 ; Chen, Wenxuan 1 ;

作者机构： 1.Zhejiang Acad Agr Sci, Food Sci Inst, Inst Agr Equipment, Inst Digital Agr, Hangzhou 310021, Peoples R China

2.Zhejiang A&F Univ, Coll Food & Hlth, Hangzhou 311300, Peoples R China

3.198 Shiqiao St, Hangzhou, Peoples R China

关键词： Dried squid; Quality assessment; Kolmogorov-Arnold network; Wavelength selection; Hyperspectral imaging

期刊名称：JOURNAL OF FOOD COMPOSITION AND ANALYSIS （ 影响因子：4.6； 五年影响因子：4.6 ）

ISSN： 0889-1575

年卷期： 2025 年 148 卷

页码：

收录情况： SCI

摘要： Given that dried squid is a highly regarded marine product in Oriental countries, the global food industry requires a swift and noninvasive quality assessment of this product. The current study therefore uses visible-nearinfrared (VIS-NIR) hyperspectral imaging and deep learning (DL) methodologies. We acquired and preprocessed VIS-NIR (400-1000 nm) hyperspectral reflectance images of 93 dried squid samples. Important wavelengths were selected using competitive adaptive reweighted sampling, principal component analysis, and the successive projections algorithm. Based on a Kolmogorov-Arnold network (KAN), we introduce a one-dimensional, KAN convolutional neural network (1D-KAN-CNN) for nondestructive measurements of fat, protein, and total volatile basic nitrogen. To compare model accuracy, the processed data were input into three machine-learning models (partial least squares regression, least-squares support vector machine, and random forest regression) and three state-of-the-art DL models (one-dimensional convolutional neural network, long short-term memory, and back propagation artificial neural network). The findings indicate that, although all DL models provide highly accurate estimates for fat, protein, and total volatile basic nitrogen in squid, the 1D-KAN-CNN outperforms the others, achieving a coefficient of determination R2 = 0.89-0.93, a root mean square error of 0.06-3.07, and a performance-to-interquartile range ratio of 2.93-3.69. These findings demonstrate the efficacy of VIS-NIR hyperspectral imaging combined with a 1D-KAN-CNN model for rapid, nondestructive, and reasonably accurate quantitative measurements of chemical content in dried squid.

  • 相关文献
作者其他论文 更多>>
置顶
购物车
反馈

© 京ICP备09089781号-29 主办单位：浙江省农业科学院

学术资源： 中国农业科学院 农业专业知识服务系统 农科联盟资源共建共享平台 中国农业科技文献与信息服务平台 中国农业期刊集成服务平台