食品科学2026，Vol.47Issue(6)：342-350,9.DOI:10.7506/spkx1002-6630-20250926-218
基于高光谱成像与深度学习融合技术的桑黄产地溯源和栽培模式识别
Origin Traceability and Cultivation Mode Identification of Phellinus linteus Based on Hyperspectral Imaging Combined with Deep Learning
摘要
Abstract
To achieve rapid,non-destructive,and accurate traceability of Phellinus linteus samples,this study compared the application of dual-band hyperspectral imaging in the visible-near-infrared(400-1 000 nm)and short-wave infrared(900-1 700 nm)regions combined with deep learning algorithms to construct a model for rapid identification of the origin and cultivation mode of P.linteus.Six preprocessing techniques were compared,including first derivative,second derivative,multiplicative scatter correction(MSC),standard normal variate(SNV),Savitzky-Golay smoothing,and detrending,along with three feature selection algorithms:successive projections algorithm(SPA),competitive adaptive reweighted sampling(CARS),and uninformative variables elimination(UVE),as well as three deep learning algorithms:convolutional neural network(CNN),back propagation neural network(BPNN),and radial basis function neural network(RBFNN).The optimal algorithm combination was determined through these comparisons.The results showed that the short-wave infrared band provided significantly higher identification accuracy than the visible-near infrared band.Among the three deep learning algorithms,the CNN model demonstrated the best classification capacity.Specifically,for P.linteus origin traceability,the SNV-UVE-CNN model in the 900–1 700 nm range exhibited the best performance,with a classification accuracy of 99.36％on the test set.For cultivation mode recognition,the MSC-CNN model in the 900–1 700 nm range performed optimally,with a classification accuracy of 97.44％on the test set.Additionally,t-distributed stochastic neighbor embedding(t-SNE)was employed to visualize the deep features extracted by the model.The findings demonstrated the advantages of the 900–1 700 nm spectral range and confirmed the suitability of the CNN model for the identification of the origin and cultivation mode of P.linteus.This study provides a pivotal theoretical foundation and essential data support for the development of intelligent rapid detection systems and the advancement of portable detection technologies.关键词
桑黄/高光谱成像/产地溯源/栽培模式识别/卷积神经网络Key words
Phellinus linteus/hyperspectral imaging/origin traceability/cultivation mode identification/convolutional neural network分类
信息技术与安全科学引用本文复制引用
史婉荣,沈书玥,王沁,李正鹏,李婷婷..基于高光谱成像与深度学习融合技术的桑黄产地溯源和栽培模式识别[J].食品科学,2026,47(6):342-350,9.基金项目
上海市东方英才青年项目 （）
上海市科委扬帆计划项目(21YF1418900) （21YF1418900）
