A geographical traceability method for Lanmaoa asiatica mushrooms from 20 township-level geographical origins by near infrared spectroscopy and ResNet image analysis techniques

Abstract

Food authenticity and traceability and climate change are key scientific issues that must be addressed in response to the food crisis in 2050. Lanmaoa asiatica mushroom is an expensive and nutritious fungi-based diets resource, it is necessary to identify its geographical origin and explore the impact of the climate on it. Thus, the purpose of this study is to establish a fast and accurate geographical traceability model based on L. asiatica mushrooms chemical information collected by near-infrared spectroscopy (NIRS) technology, and screen out key climate variables by competitive adaptive reweighted sampling (CARS) algorithm. Based on the NIRS information of L. asiatica mushrooms, two-dimensional correlation spectroscopy (2D-COS) images were generated and a residual neural network (ResNet) image recognition model was established to identify the geographical origin of L. asiatica mushrooms. The accuracy of training set and test set of ResNet model is 100%, and the loss value is 0.052, which indicates that the model is effective. In addition, the CARS algorithm was used to select the feature variables from 105 climate variables. Four important variables (February, March, and April precipitation and January minimum temperature) related to NIRS difference of L. asiatica mushroom were obtained by CARS algorithm. The results can provide a fast and accurate method for food authenticity and traceability research, and provide an innovative idea for screening key climate factors.