Human Gait Recognition Algorithm Research

Gait recognition is an emerging biometric recognition technology.Gait recognition is different from other biometric technologies in that it is easy to collect,does not require target cooperation,and is difficult to disguise.It can overcome the shortcomings of face,fingerprint and other recognition methods in certain scenarios,and has good application prospects.At present,gait recognition still faces many problems,such as susceptibility to changes in angles,changes in clothing,and changes in carrying objects.In view of the influence of these changes on gait recognition,this article has launched an in-depth study on the gait recognition algorithm.The specific content is as follows:The arrangement contour sequence diagram can fully save the space information and time information.In order to make full use of this advantage,GaitSet is used as the gait recognition algorithm.Due to the limitations of traditional computer vision algorithms in real surveillance videos,the instance segmentation algorithm MASK R-CNN is used to obtain the gait contour sequence diagram.Finally,the two algorithms are combined to realize pedestrian identification.The experimental results show that in the self-built gait contour sequence atlas,the identity of the subject can be successfully identified.The gait energy map is selected as the gait feature template for cross-view gait recognition.Aiming at the problem of cross-perspective recognition,a dual-identity discriminant generation confrontation network is proposed to transform the gait energy map.The loss functions of the two discriminators of the network are different,so that it can reduce intra-class variation,increase inter-class variation,and increase the potential information difference between generated pictures corresponding to different identities.Tested on the benchmark data set,the experimental results of this method are higher,and the effectiveness of the method is verified in the self-built dataset.