模式识别(心理学)
环境科学
数学
人工智能
计算机科学
作者
Chunlu Liu,Tao Shen,Furong Xu,Yuanzhong Wang
标识
DOI:10.1016/j.indcrop.2022.115430
摘要
As an important resource in many prescriptions, the geographical origins of Gentiana rigescens Franch. influences its chemical characteristics, quality and price greatly. Hence, a simple and rapid method for the correct classification and identification of the geographical origins of G. rigescens is of significance. In this work, marker components of iridoids were measured by high performance liquid chromatography (HPLC) and were applied as a reference to characterize chemical profiles of samples from different geographical origins. The effects of climate factors on the content differences of G. rigescens were examined by correlation analysis. Afterward, a novel two-dimensional correlation spectroscopy (2DCOS) images acquired based on Fourier transform infrared (FT-IR) spectroscopy was proposed combined to deep learning to identify geographical origins of G. rigescens. Through analyzing the iridoid components of G. rigescens, which discovered that there were significant differences in its five marker components. In addition, the marker components of gentiopicroside based on Northwestern Yunnan (DXB) were higher, and the climate environment of low temperature, temperate, and high precipitation was more suitable for the cultivation and growth of G. rigescens. In the residual convolutional neural network (ResNet), the train set and test set accuracy of synchronous 2DCOS images for the feature bands (1800–400 cm-1) was 100%, and the external validation set of all samples was correctly identified. The results indicated the synchronous 2DCOS images of feature bands were suitable for the correct identification of the geographic origin of G. rigescens, and it reduced the amount of computation and time, and saved computing resources. This study provided a powerful and useful tool for the cultivation and geographical origins identification of G. rigescens.
科研通智能强力驱动
Strongly Powered by AbleSci AI