Research On Identification Method Of Traditional Chinese Medicine Based On Deep Learning

After thousands of years of inheritance,traditional Chinese medicine(TCM)has become an important method of clinical treatment of traditional Chinese medicine.However,they have a wide variety of different quality levels,so they need to rely on effective identification methods.Among many identification methods,traditional identification methods mainly rely on artificial subjective feelings and experience.Chemical testing or instrumental analysis methods require a lot of professional equipment and identification time.Therefore,they have high cost and poor operability in actual application process.With the development of artificial intelligence technology,the TCM Informatization Development Plan which formulated by the State Administration of Traditional Chinese Medicine clearly states that informatization methods should be used to drive the modernization of TCM.The identification method of Chinese medicinal materials and the deep learning method based on big data began to combine.However,most of the current methods ignore the effect of image background characteristics,and the network model is too complex,leading to poor recognition performance and migratability.In addition,in the actual application scenarios,especially for the identification in the law enforcement process,the identification results are not quantifiable and cannot provide a corresponding legal basis.Finally,in the current field of Chinese medicine identification,there is no large-scale standard image data set available for use.Therefore,in response to the above problems,this paper conducts a study on the identification of Chinese medicinal materials based on deep learning.The main contents of the work are as follows:(1)In this paper,images of Chinese medicinal materials are collected through self-developed image acquisition equipment,and a standard Chinese medicinal material image data set with tags is constructed through standardized processing,which provides a data basis for subsequent intelligent identification research of Chinese medicinal materials;(2)Taking into account factors such as background characteristics,recognition performance and speed,this paper proposes a lightweight Chinese medicine recognition network model based on attention mechanism(Attention-TCM-Net).This model strengthens the attention to the characteristics of traditional Chinese medicine by introducing channel attention mechanism and spatial attention mechanism.At the same time,this article improves the mobile inverted bottleneck convolution module of the model,which improves the accuracy of traditional Chinese medicine recognition while ensuring a lightweight design;(3)This paper proposes a traditional Chinese medicine feature description network model based on adaptive attention mechanism.Based on the standard description definitions of Chinese medicine traits in the Chinese Pharmacopoeia,this article innovatively proposes an intelligent description method for Chinese medicine traits using deep learning models.This method uses a scoring system to complete the identification of Chinese medicine categories through feature description.This method provides a new technical route for the intelligent analysis of the traits of traditional Chinese medicine.It also provides a new solution to problems such as the lack of legal basis and related explanations for the existing appraisal technical results;(4)Applying the above two research results to the identification scene of Chinese medicine,this paper also designs and implements an intelligent Chinese medicine identification system,which makes a useful exploration of the application of computer information technology in the evaluation of Chinese medicine.