外文摘要：Recent studies on the origin of food rarely focus on the source of counties or even small-scale districts, but traceability of small-scale districts of food is the research trend and difficulty for future research. The geographical origin of Boletus bainiugan is of great significance to its safety and economic value. The aim of this study was to provide a new way for the traceability of B. bainiugan in small-scale districts using synchronous two-dimensional correlation spectroscopy (2D-COS) images combined with residual convolutional neural network (ResNet) model. In our study, 550 wild-grown B. bainiugan mushrooms were collected in 28 small-scale districts and their fourier transform near infrared (FT-NIR) were collected. First, samples from nine regions in Yunnan province were identified, and then samples from Chuxiong, Kunming and Yuxi were identified. Most of these models had a 100% accuracy rate and were suitable for identifying the geographical origin of B. bainiugan. In addition, the combination of 7000-4000 cm-1 and 15 contour lines achieves the most reliable accuracy in continuous classification model, with 100% accuracy rate in the training set, the test set and the external verification set, and the loss values were all close to zero (0.018, 0.069, 0.040 and 0.028). The results indicated that the synchronized two-dimensional correlation spectroscopy images combined with the ResNet model had excellent discriminant ability, and this analysis method provided a possibility for food identification in small-scale districts.
外文关键词：ResNet;Geographical origin;Small-scale districts;Boletus bainiugan;FT-NIR;2D-COS
作者：Li, Jieqing;Liu, Honggao;Chen, Xiong;Wang, YuanZhong
作者单位：Yunnan Acad Agr Sci;Yunnan Agr Univ;Zhaotong Univ
期刊名称：JOURNAL OF FOOD MEASUREMENT AND CHARACTERIZATION
期刊影响因子：0.0
出版年份：2024
出版刊次：18（5）
原文传递申请：江苏省科技资源（工程技术文献）统筹服务平台