Gait Recognition Based On Convolution Neural Network And Class Energy Image

Gait recognition is a highly potential identification method and has become a research hotspot for long-distance and non-contact identification.In recent years,deep learning has taken off in the field of pattern recognition and artificial intelligence,and convolutional neural networks,as an important algorithm in the field of deep learning,have strong big data processing capabilities and learning capabilities,has made remarkable achievements in image classification,is widely used in multi class object recognition.So this paper will do multi-directional research on gait recognition based on convolutional neural networks and multi-class energy images.The main work is:1.In this paper,we will implement various energy images,through comparative analysis,we select three kinds of energy images that fully contain the dynamic,static,and temporal characteristics of the gait.According to the data expansion method designed in this paper,the database of the three energy images used in the experiment was established..2.For the problem that the existing gait recognition research based on energy image does not take into account the fact that there are different perspectives and different forms of gait in real life.,this paper uses AlexNet and GoogLeNet model as the basis for adjustment,and performs cross-view and cross-form gait recognition experiments..3.For the problem that the differences between samples will affect the recognition of cross-fprm,the energy image segmentation method is used,and a three-way convolutional network model is constructed using the Merge fusion layer to improve the cross-form gait recognition effect.4.For the problem that a single energy image cannot fully include gait features,based on the RGB three-channel principle and the three-way convolutional network model,a fusion method for the data level of the three energy images is realized,and compare the gait recognition effects of different methods to verify the effectiveness of the three energy image information fusion methods.In this paper,the gait recognition method of convolutional neural networks and class energy images is used to complete the cross-view and cross-form gait recognition experiments of single energy image and multiple energy image information fusion,and the quantitative experimental results are given.The experiment verifies the effectiveness of data-level fusion and feature-level fusion methods for three energy images,and the feasibility of convolutional neural network for gait recognition.