The Research Of Convolutional Neural Network And The Application Of Image Recognition

In recent years,deep learning has gradually become known to people.Deep learning is a deeper neural network model based on neural network.Due to the continuous improvement of computer hardware facilities,computer technology has been rapidly improved and the development of deep learning has been promoted.This technology has been widely used in speech recognition,image recognition,digital recognition and other fields,thus promoting the development of science and technology.Among them,convolutional neural network is a typical representative of deep learning.Currently,most existing image recognition methods are machine learning methods,which not only require a lot of work and unsatisfactory recognition effect,but also require a lot of training time.Convolutional neural networks as cranial nerve system,it can solve the problems arising in the process of feature extraction,simultaneously has the characteristics of the weights of sharing,it highlights the convolution of the neural network advantages,with strong adaptability and learning ability,so they are widely used in image recognition,target tracking and object detection,and many other fields.In this paper,the theoretical knowledge and feature structure of the convolutional neural network are studied in detail,and the two processes of training and testing in the convolutional neural network model are deeply analyzed.On this basis,this paper combines the convolutional neural network and image recognition,and proposes an image recognition method based on the convolutional neural network.This paper first introduces the background and significance of the study of convolutional neural network and the research status and development of the subject at home and abroad,and makes a detailed understanding of the current image recognition technology.This paper focuses on the application of convolutional neural network in image recognition.Considering that the original image is the learning sample of the convolutional neural network model,the model only receives the spatial features of the image,but loses the frequency features.The image is regarded as a two-dimensional discrete signal.The frequency domain diagram of the image reflects the distribution of energy on the corresponding basis.This paper will discuss the help of frequency domain information to image recognition.This article is based on simple convolution neural network was improved,through Fourier transform and discrete wavelet transform processing the original image for image frequency domain information,and make harmony spatial and frequency domain characteristics convolution neural network learning,doing experiments on MNIST handwritten digital data sets,through Accuracy,Precision and Recall and F1-score a total of four evaluation methods were evaluated,the model based on Fourier transform convolution neural network based on discrete wavelet transform with improved convolution neural network on the 4 indexes were obtained.