Research And Implementation Of Chinese Staple Food Identification Technology Based On Convolutional Neural Network

With the rapid development of modern science and technology,artificial intelligence has formed a close relationship with people's daily life.Clothing,food,housing and transportation are the components of human daily life,in which diet is an indispensable part.Therefore,in the field of artificial intelligence,many researches related to diet have become popular.As food recognition is the basis of other food related research,this topic is a research field of great concern.However,most of the research on food identification technology is focused on foreign food and cuisine,and little on Chinese food identification technology.Therefore,this paper studies an image recognition technology for Chinese staple food species recognition,aiming to apply Chinese staple food recognition technology to people's daily life.In this paper,a method of Chinese staple food image recognition based on convolution neural network is proposed.In view of the possible interference in the data,an adaptive threshold edge detection method is proposed to process the data and complete the training in the adjusted neural network model.The new model is applied to the development of mobile phone application for Chinese staple food recognition through transformation.It mainly includes the following work:(1)Firstly,the Chinese staple food recognition data set is established,and then the data set is enhanced according to the difference of the distribution of various data in the data set.By flipping and adjusting the brightness of the original data image,the amount of all kinds of data is approximately balanced,and the size of the data set is increased.At the same time,improving the quality of the data set is also an improvement of the generalization ability of the training model;(2)Due to the diversity and uncertainty of actual data acquisition,an improved adaptive threshold detection method is proposed to deal with the data set in order to prevent the background interference contained in the input from affecting the training of neural network.This method largely eliminates the background interference and improves the generalization ability and accuracy of the model for Chinese staple food identification;(3)Considering the size of the data set used and the actual experimental conditions,in order to pursue higher experimental efficiency,the concept V3 model with relatively few parameters is used to complete the staple food identification task,and the network structure is modified to make it suitable for the Chinese staple food identification task in this paper.The experimental results show that the adaptive threshold edge detection method is feasible;(4)The training model is transformed into a model suitable for mobile platform,and the application of Chinese staple food recognition is developed on the mobile platform,and the application test is tested and satisfactory results are obtained.