Image Retrieval Algorithm Based On Convolutional Neural Network

Recently,with the emergence of low-cost cameras and smart phones,the phenomenon of people capturing images is becoming more and more common,which makes the number of online images increase dramatically.How to quickly and efficiently retrieve images from large unorganized image databases to meet user requirements is a very challenging issue.Traditional retrieval methods are inefficient because of the low expression ability,which is not conducive to large-scale image retrieval.Especially with the arrival of the era of big data,artificial intelligence joins the research of image retrieval and shows good performance.Therefore,this paper makes a deep discussion on the image retrieval technology based on convolutional neural network.The main work is as follows:A new image retrieval algorithm based on convolution neural network and principal component analysis is proposed.By designing a new network framework,we can learn probability based semantic-level similarity and feature-level similarity simultaneously.For convolutional neural networks,the output of the last fully connected layer is divided into two ways.Some of them work in n-way softmax classifier and represent the semantic information of the image.The other part is the feature information of the image.PrincipalComponent Analysis is added to convolutional neural network to obtain dimensionality reduction information of deep features.Principal Component Analysis aims to reduce the dimension of deep features while reducing the loss of information by considering the correlation between features.Finally,the distance function is used to calculate the distance between the target image and the database image to realize the retrieval.A new image retrieval algorithm based on convolution neural network and hash function is proposed.In feature extraction of image retrieval,convolution neural network is used to train and learn image features,and hash layer is added to the model of convolution network to quantify image features.By connecting Bloom filter and filtering part of the query image,the retrieval accuracy is improved and the retrieval time is reduced.Experimental results on image datasets such as ImageNet-1000,Oxford 5K,Holidays and indicate that the proposed algorithm can effectively enhance the expression ability of image features and improve the retrieval performance,which is superior to the current mainstream methods.