Image Retrieval Technology Based On Deep Learning

At the end of the 20 th century,with the popularity of mobile devices such as mobile phones and the rise of various websites,the number of images has grown very rapidly.Under this circumstance,how to retrieve the required images from a large number of images has always been a hot issue in the field of image retrieval.The extraction of image features and the similarity comparison between images are very important in the field of image retrieval.In text-based and content-based image retrieval technology,it is difficult to fully understand the real intention of the query in low-level representations of image content.Deep learning is one of the most important breakthroughs in the field of artificial intelligence in recent years and many achievements have been made in image tracking and image recognition.Therefore,this paper combines Deep Learning with image retrieval to obtain the feature representation of the image,then it studies the retrieval effect based on the feature.Not only has the strong theoretical significance but also has the highly practical value.Firstly,the paper introduces the main contents of deep convolution neural network,traditional image retrieval method and image feature extraction method.On this basis,the method of processing the features extracted by deep convolution neural network is studied,and the similarity of images is compared by distance metric learning algorithm,and an image retrieval system is designed and implemented.The main study is as follows:(1)The dimensional disaster problem caused by high-dimensional features of images extracted b y deep convolutional neural networks is studied.The PCA dimension reduction method is used to d eal with the dimension reduction of DCNN features.On the basis of dimension reduction,an impro ved PCA dimension reduction binarization(PCA-binarization)method is used to process the feature s.The binary hash code is used in image retrieval to improve the efficiency of image retrieval.(2)In order to improve the accuracy of image retrieval,this paper applies the KISS distance measure learning algorithm to measure the similarity between two images.KISS distance metric weights all dimensions of image features and can obtain better similarity measurement.By comparing and analyzing the retrieval performance of this algorithm with different similarity measure functions,distance metric learning can improve the precision of image retrieval.(3)An image retrieval system based on deep convolutional neural network is designed and implemented.The system can return the query result according to the image to be retrieved input by the user,and the results are highly similar to the images to be retrieved and have high practicability.