DEEP LEARNING-BASED GAIT RECOGNITION METHOD

摘要

Disclosed is a deep learning-based gait recognition method. The method comprises: identifying, according to a gait of a person in a video, the identity of the person through a dual-channel convolutional neural network sharing a weight and by means of a powerful leaning capability of the deep-learning convolutional neural network. The method has strong robustness for a gait change crossing a great visual angle, and effectively solves the problem of the low accuracy of an existing gait recognition technology in dealing with the cross visual angle gait recognition. The method can be widely used in scenarios equipped with video monitoring, such as security monitoring in an airport or a supermarket, personal recognition, and criminal detection.