Gait Recognition Technology Research Based On Video Sequence

With the increasing demand for intelligent monitoring and public security in security sensitive occasions such as public transport,banks and large gatherings,the subject of identity judgment through remote and video monitoring has attracted the attention of numerous scholars and relevant institutions.Gait recognition technology,as the second generation of biometric technology,uses the subtle differences in the pace to identify the identity.It can complete the identification task in the case of a long distance from the target and low video resolution.It has great academic significance and broad prospects for development.This paper focuses on the gait recognition technology applied in the video sequence,and studies the following aspects from the collection,preprocessing,feature extraction and gait recognition algorithm of the gait video sequence:Firstly,a layered moving target detection algorithm based on adaptive texture features and codebook model is proposed.The region level moving target detection is carried out by adaptive siltp operator,and the pixel level moving target detection is carried out by codebook model.Different granularity layered detection algorithm is adopted to avoid the secondary calculation of pixel value.The algorithm is tested and evaluated on the data set,which proves that it has good robustness in complex environment.Secondly,in the research of gait recognition algorithm,a gait recognition algorithm based on state space is proposed.The method of conditional random field is used to recognize gait features,and the design of conditional random field model and the optimization of parameter learning are carried out.Finally,the average accuracy of the algorithm is 97.56% in the panoramic experiment platform,and the AUC value is 0.881 through ROC curve calculation,which shows that the algorithm has a good recognition accuracy.In addition to the above algorithm,a gait recognition algorithm based on artificial neural network is proposed.At present,artificial neural network is a popular direction of gait recognition algorithm.In this paper,CNN is used to extract spatial features of gait video sequence,LSTM network to extract temporal and spatial features of gait video sequence.The average accuracy of the algorithm is 97.85% on the panoramic experiment platform,and the AUC value calculated by ROC curve is 0.890,which shows that the algorithm has a good recognition effect.Finally,through the international mainstream gait data set,the two gait recognition algorithms are compared with several mainstream gait recognition algorithms.The results show that the accuracy of the two algorithms are 97.12% and 97.64% respectively,which are superior to other algorithms.In addition,this paper also carries out the experiment of the influence of different interference factors on the recognition effect of the algorithm.From the results,it can be seen that the accuracy of the two algorithms proposed in this paper decreases by no more than 10% under the interference of light,clothing occlusion and different travel speed,which shows that the algorithm has good robustness,which provides a firm foundation for the application of the algorithm in the actual working environment.