Discussion On The Digital Expression Of "Shaped Color" Of The Properties Of TCM Based On Artificial Intelligence Technology

Prepared slices of Chinese crude drugs are processed by Traditional Chinese medicine(TCM),such as net system,cutting and cannoning.The efficacy of TCM is often changed during the processing.As a source of clinical prescription drugs,it is especially important to ensure its quality.According to the statistics of literature,the current quality evaluation techniques of TCM mainly focus on physical and chemical identification.Which is to achieve the purpose of quality control for prepared slices by detecting the indicator components or characteristic substances.That has the advantages of high sensitivity and precision,and high reliability.However,the TCM has the characteristics of “multi-component,multi-target,multi-effect”.It is difficult to reflect the overall quality of TCM by simply controlling one or several chemical components.As a summary of experience theory,the properties evaluation can objectively reflect the quality of prepared slices to a certain extent.However,it is mostly based on the experience of professionals.Although it is simple and easy to implement,the results are inflenced to subjective factors and external environment.And the objectivity and consistency are difficult to guarantee.With the development and application of big data and artificial intelligence(AI),the State Administration of TCM has formulated the "13th Five-Year Plan for the Development of Informatization of TCM",which clearly proposes to use digitize technology to drive the progress of modernization of TCM.The inheritance and innovation of TCM is becoming more and more important,and the application of new technology in TCM is imperative.By simulating the functional properties of human brain for decision-making,it is known that the identification of properties of TCM is consistent with the theory of AI.And it is feasible to objectively,digitize and intelligentize the characteristics of TCM.Based on this,it takes the identification of prepared slices as a breakthrough,and takes the four kinds of TCM including Fritillariae Cirrhosae Bulbus,Pinelliae Rhizoma,Crtaegi Fructus and Zanthoxyli Pericarpium as the research object,and the deep learning algorithm are used to realize the extraction of data from the shallow surface features to the deep semantic features.The AI model of properties of TCM is constructed,and it is compared with the image processing algorithm.The intelligent assistant evaluation of properties of prepared slices is completed.The main contents and results of reasrech are shown as follows:(1)The construction of the “shape color” database of prepared slices including Fritillariae Cirrhosae Bulbus,Pinelliae Rhizoma,Crtaegi Fructus and Zanthoxyli Pericarpium.All slices are in accordance with the Pharmacopoeia and Local evaluation standards of medicinal materials,which are identified by the old pharmacy and processing experts.The "shape" data of TCM are obtained by the machine vision technology,mobile equipment,which to carry out data cleaning,expansion preprocessing.The database of TCM,which is suitable for machine learning is build.(2)The image classification research of TCM decoction pieces: the 60 pieces of 780 image of different types of pieces are collected,and the multi-dimensional features of color,shape and texture are extracted as input matrix.After preprocessing operations such as image graying and binarization,the 13 different types of decoction pieces were used as the output matrix.The SVM and ANN classifier were selected to construct the intelligent identification model based on supervised machine learning method.According to the experimental results,the SVM has higher accuracy.(3)The research on classification of TCM: Based on the above-mentioned database,the YOLO_v2 is used to detect and classify multiple types of prepared pieces,and the data is calibrated by Label Img.The data set is listed as Fritillariae Cirrhosae Bulbus,Pinelliae Rhizoma,Crtaegi Fructus and Zanthoxyli Pericarpium 4 types.At the same time,the normalize data is adjusted to 416×416 as the input of network,and 50 as the output tensor.The multiple types of prepared pieces are detected by adjusted the network parameters.The experimental results show the method can better complete the detection of prepared pieces,so as to realize the differentiation of large types of pieces.(4)The research on intelligent identification of prepared pieces: According to the processing results in the previous chapter,the convolutional neural network is used to extract feature and identificate intelligently for many kinds of prepared pieces.All data are normalized to 256×256 as the input of neural network,and the number of 13 as the output.The 80% is used as the training set,20% is used as the test set data.The network parameters are adjusted to identificate intelligently.The experimental results show that the highest accuracy of the test set is 96.46%,which proves that the method can solve the classification of many kinds of prepared pieces.In summary,the properites of prepared slices of Chinese crude drugs as a breakthrough point in this paper.The deep learning algorithm are applied to research the quality evaluation of TCM.Through the construction of database,preprocessing,detection,identification and classification.The research is confirmed that the AI can effectively realize the intelligent assistant evaluation of the properites of prepared slices of Chinese crude drugs.