Research On Key Technologies Of Content Based Image Retrieval

Content-based image retrieval is a retrieval system of finding images based on an image.It can retrieve related images in an image database on the basis of the internal implicit semantic information of an image.For existent content-based image retrieval technology,the system will extract the visual features for image retrieval.These visual features include color,texture,shape,key points and so on.However,the feature descriptors designed by people often do not have the characteristics of generality and universality,one descriptor may have a good expression for a particular object,but do not do well for others.In order to solve this problem,this thesis proposes the use of deep learning techniques to extract the image feature automatically,and converts these features into local sensitive hash table,to accelerate the speed of image retrieval.The main work of this thesis is as follows:(1)Make use of AlexNet network to extract the image feature automatically.First,use the large image dataset to pretrain the network to capture feature information of the general images.Second,on this foundation,train the network again with small goal dataset,to get image features of matching with goal dataset automatically.Third,treat the output of the Fc7 layer in the trained network as the image feature.(2)For features of extraction,index them by p-stable Locality Sensitive Hashing,and store them into hashing indexes,to accelerate the speed of image retrieval of resembling goal image.(3)Design comparative experiments and compare with image retrieval system based on color moment and so forth.It indicates the proposed AlexNet-based image retrieval system of this thesis have obvious advantages in precision and recall,thereby the problem of conventional feature extraction which do not have universality and robustness can be solved.