Research On Image Segmentation And Case Analysis Algorithm Of Tongue Fur By Using Convolutional Neural Network

With the development of artificial intelligence technology becoming more and more mature,there is a shadow of artificial intelligence both in people's life and learning.In the process of technology development,hardware upgrades are very rapid.In recent years,deep learning has begun to take a new step with the promotion of hardware.Image segmentation is a direction of artificial intelligence,and the accuracy of the classification result is directly affected by the segmentation result,so segmentation plays a crucial role.The diagnosis of tongue on Chinese medicine is based on the "seeing",which is a traditional method.With the development of modern technology,intelligentization and objectivity are essential for Chinese medicine.Modern intelligent technologies are of great assistance to TCM,and deep learning can assist in the diagnosis of TCM and enhance its effectiveness.Promote the development of Chinese medicine.This topic mainly studies the segmentation of tongue coating and classifies it for segmentation.Firstly,the working principle of the segmentation network and related algorithms are explained.Secondly,the full convolutional network is used to segment the tongue fur and use pixel-level annotation images to use the upsampling and skip structure in the full convolutional network.Resume.The segmentation process is based on the theory of VGG16.A network architecture called TonNet is proposed.Then the working principle of the classification network and related algorithms are described.After the segmentation is completed,the marked data sets are further classified,thus making the network fully functional.The use of this also makes the result more accurate.In the end,the experimental results of segmentation and classification are used to compare the experimental results of two different networks.From the experimental data,it can be seen that the selected TonNet segmentation and classification results perform better than VGG16 segmentation and LeNet5 classification.