Image retrieval method based on multi-feature and multi-relationship

Abstract

The invention discloses an image retrieval method based on multiple features and multiple relations. The image retrieval method based on the multiple features and the multiple relations is characterized by being carried out according to the following steps of 1, extracting the color feature, the texture feature, the shape feature and the bag-of-words feature of an image; 2, fusing the extracted features of different types into one comprehensive feature; 3, inputting the comprehensive feature into a convolutional neural network to obtain a better feature expression; 4, carrying out k-mean clustering on a result obtained by the convolutional neural network, and using the centers of all clusters as different relational nerve cells; 5, constructing a target equation; 6, constructing a similarity equation; 7, carrying out iterative updating on each relational nerve cell by utilizing a BP algorithm. By using the image retrieval method based on the multiple features and the multiple relations, the image retrieval can be carried out better; the robustness is enhanced.

Description

technical field [0001] The invention belongs to the field of image retrieval, and mainly relates to an image retrieval method based on multiple features and multiple relationships Background technique [0002] With the development of Internet technology, people have become accustomed to searching for pictures on the Internet. However, at present, mainstream websites such as Baidu and Google use text tags manually given in images for retrieval. If no tags are added to the images, they cannot be retrieved. Therefore, how to use images to retrieve images is one of the hot topics in the field of Internet image retrieval. [0003] In response to the above problems, researchers have proposed some methods, but because images have different features and the complexity of image relationships, using images to retrieve images has always puzzled researchers. Recently, an article titled "Neighborhood Discriminant Hashing of large scale image retrieval" was published on IEEE TRANS ON IM...

AI Technical Summary

Problems solved by technology

[0004] The present invention aims to solve the problem of unsatisfactory image retrieval results caused by ignoring the different features and relationships of images in the current image retrieval method, and proposes an image retrieval method based on multiple features and multiple relationships, in order to achieve comprehensive Consider different features of images, as well as different relationships between images, to improve the effect of image retrieval, thereby enhancing the robustness of image retrieval

Images