Cross angle of view gait recognition method based on two-dimensional coupling margin Fisher analysis

Abstract

The invention provides a cross angle of view gait recognition method based on two-dimensional coupling margin Fisher analysis. According to the invention, the characteristics of a gait energy graph at different angles of view and the superiority of coupling metric learning in the aspect of cross-domain biometric feature recognition are combined; two-dimensional coupling margin Fisher analysis is provided; data difference of the cross angle of view gait energy graph in matrix space is weakened; local relation among samples is kept; the intra-class divergence is the largest; the intra-class divergence is the smallest; and the cross angle of view gait recognition performance is greatly improved.

Description

technical field [0001] The invention relates to a cross-view gait recognition method based on two-dimensional coupling margin Fisher analysis, which belongs to the field of pattern recognition and machine learning. technical background [0002] With the rapid development of computer computing and storage technology, biometric identification technology based on computer vision is widely used in the fields of business, security, medicine, military and entertainment. In recent years, biometric identification technology under controllable conditions has become more and more mature, while non-controllable conditions such as long-distance, non-contact and low resolution pose severe challenges to traditional biometric identification technology. [0003] Gait features have the advantages of long-distance non-contact detection, not easy to camouflage and imitate, and are less affected by the external environment. However, the change of gait viewing angle will lead to differences bet...

AI Technical Summary

Problems solved by technology

In 2006, Zhao et al. proposed a multi-camera-based 3D gait tracking and recognition method, using multiple collaborative cameras for gait analysis; in 2009, Bodor et al. A registered sample view reconstructs the gait sample to be tested under any view, both of these methods require a complex collaborative multi-camera system

In 2012, Ben et al. proposed a cross-view gait recognition method based on coupled distance metric learning, which connected gait features from different perspectives by minimizing the error of similar samples from different perspectives in the common subspace, and retained samples The local information and manifold structure between them greatly improves the recognition rate of cross-domain biometric features, and also achieves good results in gait recognition. However, this method is based on vector manifold alignment. When the number of samples is small and the dimension When the number is high, it is easy to cause the "curse of dimensionality"

Existing cross-view gait recognition methods often require complex multi-camera collaboration systems or look for view-insensitive features with poor robustness

Existing cross-view gait recognition methods based on coupled metric learning are all vector manifold aligned, and need to map two-dimensional gait samples to one-dimensional vector space first, which not only destroys the structural information of the gait energy map, but also It is easy to cause the "curse of dimensionality"

Images