Identification Of Chinese Herbal Medicines Based On Deep Neural Network

Since the deep learning was put forward in 2006,it has made great achievements in the field of intelligent applications such as text,voice,image and other information carriers.Today,deep learning has been applied to more fields of intelligence,and has achieved remarkable successes in these fields.At present,people are paying more and more attention to health care in traditional Chinese medicine(TCM).Traditional Chinese medicinal materials are the important parts of TCM,but most of people are not experts in TCM and lack of the knowledge to recognize the variety of traditional Chinese medicinal materials.In the research field of Chinese medicinal materials recognition,the existing research is limited to a small amount of herbs or relys on specialized equipment,which lacks of practicality.Meanwhile,the research in this field with deep learning technology is still in the blank stage.Through the use of deep learning,it can be more intelligent to identify Chinese medicinal materials,which is beneficial to people and markets.We apply deep convolutional neural network(CNN)into the traditional Chinese medicine materials recognition task and we notice that,many materials in this task are similar,they can only be distinguished by some low-level features such as texture and edge features,etc.So we propose a new method of feature fusion,which fuse the features from low-layer and high-layer to imporve the performance.Further,the Bagging method in Ensemble learning is used to train a number of weak classifiers,and a strong classifier for Chinese medicine recognition is made by voting method.The experimental results show that the algorithm can achieve high accuracy in our 98 categories of Chinese medicinal materials dataset.Meanwhile,we implement a Chinese medicine materials identification APP,which runs smoothly and gains high accuracy with practical value.