Method for image classification and brain-imitated storage based on user group optimization

Abstract

The invention provides a method for classified storage of distributed image documents based on a brain storage mode. The method comprises the operation steps that (1) an image data expression model is established; (2) a brain-imitated distributed storage system is established; (3) images are labeled and classified automatically; and (4) label amendment is conducted. With the method, automatic labeling results of the images can be obtained through automatic labeling; and then the label amendment can be conducted based on a user's understanding of pictures. The method can also feed back similar images according to the user's demands for image types. The user can obtain classification results of the designated images and similar pictures. The user can conduct artificial intervention and amendment of the obtained classification results, so as to improve accuracy of the image classification.

Description

technical field [0001] The present invention relates to the field of computer artificial intelligence, in particular to an image classification brain-like storage method based on user group optimization, which is a brain-like distributed storage of images and an image classification storage method combining picture features with traditional classifiers . Background technique [0002] In the current information age, the growth rate of data scale far exceeds our expectations, and there are a large number of picture files. In this case, the centralized file system cannot effectively manage these image files. The traditional approach to image management is text-based retrieval. Although this method is easy to implement, it is difficult for text description images to fully express rich content. In addition, manual annotation is time-consuming and laborious. There are many kinds of images, and there are also many ways to store different image types. There is a data structure a...

AI Technical Summary

Problems solved by technology

Existing network sharing models cannot fundamentally eliminate the problem of "information islands"

[0004] In addition, the current traditional image classification and storage methods are all single-mode, in other words, the storage system is mainly responsible for access, while the classification system is only responsible for labeling the images, and finally the classification information is sent back to the storage system

This method is inconvenient for users to operate

Therefore, when the image required by the user is incorrectly classified by the system, the system will not be able to return the correct category, making the traditional image classification storage method unable to meet the user's needs

[0005] The existing image annotation classification is roughly divided into two annotation methods: manual annotation and automatic annotation. Among them, pure manual annotation requires a lot of manpower and material resources, which becomes impractical.

The automatic labeling of images can solve the problem of image labeling to a certain extent through the method of machine learning, but because the feature perception of automatic labeling is not perfect, it is difficult to reflect the concept expressed by the image, and the accuracy of automatic labeling is not high

Images