Individual gait recognition method based on trinocular vision data

Abstract

The invention provides an individual gait recognition method based on trinocular vision data. The method comprises the following steps: S1, collecting human body contour information at multiple angles; s2, extracting gait dynamic characteristics under the human body contour information of each angle to form a gait pattern library; s3, constructing a walking deep learning model at each angle according to the gait dynamic features extracted in the step S2; s4, obtaining gait dynamic characteristics under the multi-angle human body contour information to be recognized, and recognition errors based on the gait dynamic characteristics in an existing gait pattern library; and S5, according to the size of the obtained recognition error, judging the weight value of the feature vector obtained by learning of the gait deep learning model at each angle in the final classification task, and realizing gait recognition. According to the method, gait information under multiple different observation view angles is fused, and comprehensive characteristics are extracted. Therefore, the method is more suitable for the influence of complicated and changeable internal and external factors in practical application, and has higher practicability and operability.

Description

technical field [0001] The present invention relates to the technical field of pattern recognition, and more specifically, to a method for recognizing individual gait based on trinocular vision data. Background technique [0002] Although many gait recognition algorithms have emerged, most of these works rely on a specific observation angle, and only collect and analyze the gait characteristics under a single observation angle, so they are often different in different clothing, site background, bag or not, wearing The recognition rate is not high when the shoes are of different types and the pace is different. The gait characteristics collected from a single observation perspective are relatively limited, and it is impossible to comprehensively and deeply describe the internal information of the walking process, so that it is impossible to build a gait recognition system with strong robustness and strong anti-interference ability. [0003] The public date is October 23, 202...

AI Technical Summary

Problems solved by technology

This patent adopts a single observation perspective, which cannot comprehensively and deeply describe the internal information in the walking process

Images