Research On Image Retrieval Algorithm Based On BP Neural Network

With the development of multimedia technology,a large number of images are produced in the network,which makes the acquisition of image information become more and more important.How to obtain the necessary content from the mass image accurately and quickly,it involves the classification and retrieval of the image.All the time,traditional image retrieval methods are based on text retrieval.However,with the increasing number of images,text-based image retrieval has been unable to meet the needs of people's information retrieval.Content-based image retrieval technology describes the features of the image through the system,and extracts the visual features of the image to realize the image retrieval.It overcomes the shortcomings of the traditional text retrieval,such as the one-sided and subjectivity,which are brought by the manual annotation.In recent years,content-based image retrieval technology has attracted wide attention,and has become a hot research topic in this field.Firstly,this paper discusses the development of content-based image retrieval technology and research results,and studies the related technologies of image retrieval and the commonly used PCA algorithm and BP neural network algorithm in image retrieval.In this paper,a fusion algorithm is presented,which combines the HSV color histogram and SIFT features.The HSV color histogram has the characteristics of intuitive color expression and simple calculation.The SIFT feature has strong robustness to image rotation,scaling and brightness changes.HSV color histogram lacks of spatial information,SIFT feature matching method do not have a high accuracy.In view of the mentioned problems,an improved fusion algorithm is proposed in this paper.It designs overlapped sub-blocks to emphasize visual content of image center zone,and the SIFT feature measurement algorithm is improved.Then,the two improved algorithms are combined to implement image retrieval.Finally,the effectiveness of the algorithm is verified by experiments.Secondly,aiming at the shortcomings of SIFT algorithm,such as high complexity and long matching time,this paper adopts dimensionality reduction algorithm of the SIFT based on PCA.The PCA algorithm is used to reduce the dimension of the extracted SIFT features,so as to reduce the complexity of the algorithm and improve the matching efficiency.In subsequent image retrieval,BP neural network is used as image classifier.An image retrieval process based on SIFT dimension reduction and BP neural network is designed.Through statistical analysis of the retrieval results and compared with other literature,it is found that the retrieval algorithm proposed in this paper can better solve the influence of zoom,rotation and other factors on the image,and has better adaptability to the natural image.