A Multimodal Image Recognition Method Based on Low Rank and Joint Sparsity

Abstract

The invention discloses a multimodal image recognition method based on low rank and joint sparseness, belonging to the technical field of image recognition. In order to overcome the technical problem that the difference between modalities in multimodal images is greater than the difference between categories, the present invention first projects the original multimodal data into a low-rank common subspace, so that the low-rank constraints on the common subspace can be effectively The similar information between different modalities of the same category can be preserved, so that the connection between categories in the low-rank public subspace is greater, and at the same time, the image dimension can be reduced, which avoids the dimensionality disaster to a certain extent, and then through the joint sparse constraint The joint sparse representation of different modal data is obtained by means of the method, and the fused features are obtained; and then the features are classified and recognized by common classifiers to obtain the final recognition result. Aiming at the problem of multimodality, the present invention uses feature fusion to combine features of multiple modalities to obtain features that are more conducive to identification and improve identification efficiency.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and in particular relates to a multimodal image recognition technology based on low rank and joint sparseness. Background technique [0002] Image recognition technology uses computers to process and analyze images, and to classify or make meaningful judgments on objects in images. With the development of sensors, it is easy to capture multimodal image data in real life. Fusion of multimodal data can provide complementary information, which improves the recognition performance, and the scheme of fusing multimodal information has greater practical application value than the scheme based on single-modal information. Due to the differences in the imaging mechanisms of different modalities, traditional single-modal image recognition algorithms cannot handle multi-modal images, which limits the further application of image recognition. There are large differences between data from differen...

AI Technical Summary

Problems solved by technology

Since the similarity within the modality is greater than the similarity within the category, it is easy to lose modality information by directly fusing features with large differences in the above-mentioned existing multimodal feature fusion methods.

Images