Research On Identification Method Of Chinese Medicinal Materials Based On Deep Learning

The development of Chinese medicine in China has formed its complete theoretical system,and Chinese herbal medicine is an indispensable part of it.There are many kinds of Chinese herbal medicines,but the number of fake and inferior products on the market is not uncommon.Non-professionals are difficult to identify correctly,which seriously hinders the healthy development of the Chinese medicine industry.It is extremely urgent to develop and establish a systematic and authentic identification technology of Chinese herbal medicine.The traditional identification methods rely too much on manual intervention,which is subjective and costly.Secondly,the existing computer-aided identification methods are mainly based on traditional machine learning algorithms,and their identification accuracy is low and time-consuming.Aiming at the above problems,this thesis mainly studies the classification and identification algorithm of Chinese herbal medicine based on deep learning,compares it with the traditional machine learning algorithm,and optimizes the neural network and constructs an independent Chinese herbal medicine image database,which finally improves accuracy and objectivity of the classification of Chinese herbal medicines.In view of the above analysis,the main research works of this paper is as follows:1)In the field of image identification of Chinese herbal medicines,there is currently no open standard dataset that can be directly used.This paper constructs a data set of 11 groups of 11 common confusing fruit seeds and Chinese herbal medicines,with an image dataset consists of 11379 pictures of Chinese herbal medicines.At the same time,the validity of the data set was verified by experiments.2)Aiming at the problems that the traditional Chinese medicine micro-trait identification method relies too much on labor and the effect is not good,a computer-aided Chinese medicine material identification method based on guided filtering and feature extraction is proposed.Firstly,combined with guided filtering,image fusion is used to replace the traditional artificial depth of field synthesis process,and the micro-traits of Chinese herbal medicines are obtained.Secondly,according to the characteristics of the medicinal materials,the features are extracted and merged after dimension reduction to achieve feature fusion.The results show that the algorithm is superior to the algorithm based on single feature and the combination of different features,which lays a foundation for the automatic identification of Chinese herbal medicines in subtle state.3)Facing the complicated and time-consuming process of traditional Chinese herbal medicine image identification process,a classification algorithm based on deep convolutional neural network for computer-assisted Chinese herbal medicine was proposed.Firstly,image fusion is realized by introducing convolutional neural network.At the same time,in order to reduce the influence of illumination during image acquisition,the image is pre-processed by homomorphic filtering before fusion.Secondly,ZCA whitening is used to reduce the micro-traits between Chinese herbal medicines.Redundancy,feature extraction through improved convolutional neural network;finally using Softmax classifier for classification and identification.Five groups of 11 Chinese herbal medicines were collected and compared with AlexNet,ResNet18 and DSC to verify the effectiveness of the algorithm.The experimental results show that the algorithm has higher classification accuracy and meets the real-time requirements,which is more practical.