An Event-Related Social Image Search Algorithm Based on Hypergraph Model

Abstract

The invention discloses an event-related socialized image searching algorithm based on a hypergraph model. The event-related socialized image searching algorithm is characterized by comprising the following steps: 1, extracting the modal features of the socialized images; 2, respectively establishing hyperedges on any socialized image; 3, forming the hypergraph model by a hyperedge set; 4, establishing a ranking model based on the hypergraph model; 5, and adopting a convex optimization theory to obtain the optimal solution. According to the event-related socialized image searching algorithm, the socialized images of an Internet platform are related and archived through the sufficient utilization of multi-modal information in files, so that the most event-related socialized image is accurately searched.

Description

technical field [0001] The invention relates to social media retrieval, in particular to an event-related social image search algorithm based on a hypergraph model. Background technique [0002] With the rapid development of social multimedia, a large amount of media data is shared on the Internet. It has become a technical problem with practical significance to mine the potential relationship between social multimedia information and efficiently organize and manage social media. In recent years, events, as a very effective means of information management, have received more and more attention from academia. At present, there are different modal data sharing platforms on the Internet, and related images of recorded events are shared on these platforms in different formats by different users. How to find relevant multimedia documents from a large amount of social media data on the Internet so as to visually display events has become a current research trend. [0003] Some ...

AI Technical Summary

Problems solved by technology

[0005] First, in the traditional hypergraph model, the affiliation of vertices to hyperedges is represented by a simple 0 / 1, which cannot reflect the difference in the affiliation between different vertices and hyperedges.

[0006] Second, the hypergraph model is generally established by the nearest neighbor (KNN) strategy, that is, each document and its K neighbors form a hyperedge, and the selection of the neighbor parameter K directly affects the performance of the algorithm, but the neighbor parameter K is not easy to choose

[0007] Third, for different feature modes, they have different weights when representing events and document relationships, and the traditional hypergraph model does not optimize the weights of hyperedges because of the selection of uniform weights, so that the optimal search results cannot be obtained

Images