Research On Image Recognition And Classification Of Chinese Medicines In Western Long Medicine

Purpose —— With the development of Chinese medicine in recent years,Chinese herbal medicines have become an indispensable part of Chinese medicine,and people’s awareness of the importance of Chinese herbal medicines has gradually increased.However,the quality of Chinese herbal medicines on the market is uneven,the number of fake and inferior products is not uncommon,and the traditional identification method mainly relies on professionals,which has various problems such as low efficiency,strong subjectivity and high cost,and it is difficult for non-professionals to identify.Therefore,with the development of deep learning technology,introducing it into the identification method of Chinese medicinal materials will solve many of the above problems.This paper will use deep learning technology to complete the classification and identification of Chinese medicinal materials.Methods —— Due to the subjective and unstable factors of traditional Chinese herbal medicine identification methods,this paper proposes to use convolutional neural network to extract features and classify and identify Chinese herbal medicines.Use data enhancement to expand the data volume of Chinese medicinal materials,improve the image quality of Chinese medicinal materials through data normalization,grayscale and other preprocessing operations,and then build a suitable convolutional neural network model;and in order to further improve the recognition rate of the network model,A series of optimization operations are performed on the model,and the original network model is optimized by introducing global draw pooling,feature fusion and ELU activation function.Finally,the results of the experiments are compared,researched and analyzed,and the related experiments are completed using the deep learning framework Tensor Flow.Research Results —— This paper compares the results of network models with different depths.When the model has 14 layers,the classification accuracy of the model IV is the best.After comparing the accuracy of the global draw pooling,feature fusion and ELU activation function introduced into the model,the conclusion is that after the feature fusion is introduced into the network model,the recognition rate of the model is significantly improved,which can meet the needs of national daily life.Limitations of the study —— The categories of Chinese herbal medicines that can be classified are limited.This paper only studies a few specific Chinese herbal medicines,and has not completed the large-scale identification of Chinese herbal medicines,and cannot complete the identification of other Chinese herbal medicines.Practical impact —— The application of convolutional neural network technology to image recognition and classification of Chinese medicinal materials can not only improve the accuracy of classification and recognition,but also further simplify the corresponding classification and recognition process,complete the classification and recognition of Chinese medicinal materials,and obtain a model with better classification effect.It can meet people’s needs for identifying Chinese medicinal materials in daily life.Originality —— The images in the dataset are processed by normalization,grayscale,noise reduction and data enhancement,and the resolution of the images is improved.Using deep learning technology to identify and classify the processed images,and improve the proposed network model,which improves the performance of the model while improving the recognition rate.