Design And Implementation Of Content-Based Image Retrieval System

With the rapid development of information technology,people are exposed to more and more image information,and the massive image data has broadened people's horizons.The shortcomings of traditional text-based image retrieval,can no longer satisfy people's demand for image retrieval,so content-based image retrieval technology has emerged.The essence of content-based image retrieval is the extraction of image features.The image features need to reflect the semantic information of the image.In practical applications,different features should be selected for processing.Therefore,this thesis studies the image retrieval technology from the underlying visual features and high-level semantic features of the image.The main works of this thesis are as follows:1)The application of color space features in image retrieval system is studied.Firstly,the RGB color space and HSV color space are introduced.Then,the method of image retrieval based on color space features is studied.Experiments are carried out on the retrieval method using two color space features and the retrieval method combining two color space features.Finally,the retrieval results of these three methods in content-based image retrieval are analyzed through experiments.2)The application of SIFT algorithm in image retrieval system is studied.Firstly,the basic principle and implementation steps of SIFT algorithm for extracting feature descriptors are introduced.The formation of SIFT feature descriptors are analyzed from four aspects: construction scale space,detection and location of key points,determination of key points and description of key points process.The specific steps of implementing image retrieval using SIFT features are then visually described through a flow chart.Finally,the retrieval effect of SIFT feature extraction method in image retrieval is analyzed through experiments.3)The application of convolutional neural network in image retrieval system is studied.Firstly,the structure of convolutional neural network and the network model contained in convolutional neural network in recent years are introduced.Then,the pre-training model of Inception V3 is used to extract image features.Finally,the retrieval effect of the model in image retrieval is analyzed through experiments.Finally,in order to compare the performance of various feature extraction methods,the various experimental conditions in this paper use the same experimental conditions.The experimental results show that compared with the feature extraction using color space and SIFT algorithm,feature extraction using convolutional neural network has better retrieval effect.