Research And Implementation Of Chinese Herbal Pieces Recognition Algorithm Based On Deep Learning

With the improvement of people’s attention to health,the application of traditional Chinese medicine decoction pieces is more and more widely.Traditional Chinese medicine decoction pieces rely on manual operation in the production and dispensing process,which consumes a lot of manpower.And people may eat by mistake because they don’t know the name or drug properties of traditional Chinese medicine.In order to solve these problems and enhance people’s learning of traditional Chinese medicine,this thesis applies YOLOv4 target detection algorithm to the recognition of traditional Chinese medicine,and designs and implements the online recognition system of traditional Chinese medicine.In the traditional Chinese medicine Pieces recognition system,the category detection of traditional Chinese medicine pieces is the most critical step.By detecting the category,we can match the relevant information.The difficulty of category detection of traditional Chinese medicine Pieces lies in the complex background,different scales of target objects,and most traditional Chinese medicine pieces are small.Therefore,the YOLOv4 target detection algorithm is improved to improve the recognition performance of the algorithm.The Non-local attention mechanism is added to the algorithm to improve the recognition performance of the algorithm for traditional Chinese medicine decoction pieces under complex background,and the RFB module is added to improve the recognition performance of the algorithm for traditional Chinese medicine decoction pieces of different scales.The experimental results show that on the self-made 8469 pieces of traditional Chinese medicine image data set,the performance evaluation index map value measured by the improved algorithm is98.02%,which is significantly improved compared with 96.89% measured by the original algorithm.After obtaining better recognition performance,match the relevant information of the category of traditional Chinese medicine decoction pieces,including origin,appearance characteristics,main therapeutic efficacy,etc.The improved algorithm model is deployed on the wechat applet to realize the online recognition of traditional Chinese medicine decoction pieces.This thesis improves the traditional Chinese medicine decoction pieces recognition algorithm.Identifying the categories of traditional Chinese medicine decoction pieces through the improved algorithm and matching the corresponding relevant information can strengthen people’s learning of traditional Chinese medicine decoction pieces and prevent the occurrence of accidental eating.At the same time,it also provides a direction for the production and adjustment of traditional Chinese medicine decoction pieces.Finally,the algorithm is deployed in wechat applet to achieve the characteristics of no memory and use as you go.