Classification Method Based on Adaptive Deep Forest Human Gait Energy Map

Abstract

The invention discloses a classification method based on an adaptive deep forest human body gait energy map. The steps are: (1) constructing a training set and a test set; (2) constructing an adaptive deep forest model; (3) calculating an adaptive deep forest model. Classification accuracy of the adaptive deep forest model; (4) Judging whether the adaptive deep forest model meets the requirements; (5) If the accuracy and depth of the training set do not meet the requirements, expand the adaptive deep forest model and solve the custom quadratic convex optimization The problem is to obtain the enhanced features of the adaptive deep forest model in the training set, and obtain the expanded adaptive deep forest training set through the enhanced features on the training set; (6) Calculate the classification accuracy of the test set. The invention reduces the complexity of the model and the weight that the model needs to train, and at the same time reduces the sample size of the gait energy map required for training the model.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a human body gait energy map classification method based on an adaptive deep forest in the technical field of image recognition. The invention can be used to classify the gait features in the human body gait energy image GEI (GaitEnergy Image). Background technique [0002] Gait recognition technology is a biotechnology that recognizes people based on their walking posture in video sequences. Because gait recognition is non-invasive, long-distance recognition, and difficult to hide, gait recognition has broad application prospects in the fields of national public security, financial security, identity authentication, and video surveillance. [0003] Galaxy Waterdrop Technology (Beijing) Co., Ltd. proposed a gait based on deep learning in its patent document "A Gait Recognition Method Based on Deep Learning" (patent application number: 201410587758X, application pu...

AI Technical Summary

Problems solved by technology

This method has strong robustness to gait changes across large viewing angles, and effectively solves the problem of low accuracy of existing gait recognition technologies when dealing with cross-view gait recognition.

However, the disadvantage of this method is that since this method is a classification method based on a deep convolutional neural network, and a classification method based on a deep convolutional neural network requires a large number of training samples, in the case of a relatively small number of training samples Under the circumstances, the classification method of deep convolutional neural network cannot fully learn the gait features in the human gait energy map, thus reducing the feature extraction ability of deep convolutional neural network, which makes the classification method of deep convolutional neural network unsuitable for Small sample data, and the classification performance based on the deep convolutional neural network is heavily dependent on the parameter debugging process, which consumes a lot of computing resources during the training process

However, the shortcomings of this method are that the method requires the user to carry a smart terminal to obtain the acceleration data, and use the obtained acceleration data to extract the user's behavior characteristics, position characteristics and gait characteristics, which weakens the gait classification. The non-invasiveness of the method, and the method is composed of three models of pre-trained behavior recognition model, position recognition model and gait recognition model, it is impossible to integrate all the feature extraction into one model, and it is impossible to achieve end-to-end classification

Images