Small-scale districts identification of Boletus bainiugan from Yunnan province of China based on residual convolutional neural network continuous classification models

文献类型： 外文期刊

第一作者： Chen, Xiong

作者： Chen, Xiong;Liu, HongGao;Chen, Xiong;Wang, YuanZhong;Li, JieQing

作者机构：

关键词： Small-scale districts; Geographical origin; Boletus bainiugan; FT-NIR; 2D-COS; ResNet

期刊名称：JOURNAL OF FOOD MEASUREMENT AND CHARACTERIZATION （ 影响因子：3.4； 五年影响因子：3.2 ）

ISSN： 2193-4126

年卷期： 2024 年

页码：

收录情况： SCI

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.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序