Multi-view subspace clustering method and device

Abstract

The invention discloses a multi-view subspace clustering method and device, which can make full use of complementarity information among multi-view data, can ensure the clustering structure consistency of a self-representation matrix and can greatly improve the clustering performance. The method comprises following steps: (1) acquiring multi-view feature matrix of original data set; (2) for givenmulti-view data Xv, according to a low-rank self-representation subspace clustering method, performing low-rank constraint and rank structure consistency constraint on a representation coefficient matrix Zv of each piece of view data respectively to obtain a low-rank self-representation subspace clustering matrix Zv of each piece of view data; constructing a multi-view subspace clustering model based on low-rank matrix decomposition and rank structure consistency constraint; (3) for the constructed multi-view subspace clustering model, solving an optimization problem by adopting an alternatingdirection multiplier method to obtain a multi-view self-representation coefficient matrix with consistent rank structures; and (4) fusing the multi-view self-representation coefficient matrixes to obtain an incidence matrix W, and performing spectral clustering on the incidence matrix W to obtain a final clustering result.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a multi-view subspace clustering method and a multi-view subspace clustering device. Background technique [0002] Multi-view data refers to the description of the same sample by multiple data features from different perspectives. Multi-view data is becoming more and more common in many practical applications. For example, for video surveillance systems, multi-camera systems record human activities from multiple angles; Internet network news includes text, images, and audio and video; for image data, different features can be used to describe images, such as SIFT features, LBP features , HOG features, etc. Multi-view data describe the same sample from different perspectives, which can provide complementary information. Clustering analysis is an important data mining method, but traditional clustering methods cannot maximize the use of complementary information...

AI Technical Summary

Problems solved by technology

[0004] Although the clustering effect of the above methods has been greatly improved, most multi-view self-representation subspace clustering methods assume that multi-view data share the same self-representation coefficient matrix to explore consistent information in multi-view data.

Images