Single-view-angle target recognition method based on two-modal distance preserving related feature learning

Abstract

The invention discloses a single-view-angle target recognition method based on two-modal distance preserving correlation feature learning, which mainly comprises the following steps of: constructing acorrelation feature learning model between two modals by constraining recognition clustering dispersion while maximizing correlation between the modals, and further realizing recognition of a single-view-angle target by utilizing the model. The method comprises the following steps: (1) carrying out modal processing on a single-view image; (2) constructing a two-modal distance maintenance relatedfeature learning model, and carrying out optimization solution on the model to obtain low-dimensional related features; and (3) fusing the low-dimensional related features by using a parallel fusion strategy, and finally obtaining an recognition result of the single-view target by means of a nearest neighbor classifier. Compared with the prior art, the method can effectively improve the recognition precision of a single-view-angle target by means of the advantages of two-modal feature learning.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition and information fusion, and specifically relates to a single-view target recognition method based on two-modal distance-preserving correlation feature learning. Background technique [0002] Feature learning is a popular research topic in pattern recognition. According to the amount of modal data corresponding to the same target, feature learning can be divided into single-modal feature learning and multi-modal feature learning. Principal Components Analysis (PCA) and Locality Preserving Projections (LPP) are classic single-modal feature learning methods and have shown good performance in many practical applications. Compared with the single-modal feature learning method, the multi-modal feature learning method can use the advantages of multi-modal feature learning to more comprehensively grasp the internal structure of the target data. However, how to use multi-modal feature learning...

AI Technical Summary

Problems solved by technology

However, these solutions often ignore the local structure hidden in the original high-dimensional data, so it is difficult to effectively capture the geometric structure hidden in the high-dimensional data, and the original high-dimensional data contains a lot of redundant information and noise, This will further limit the recognition performance in actual image recognition

Images