Research On Gait Recognition Technology Based On Contour Features

In recent years,crimes such as illegal attacks and illegal invasions have occurred frequently.The issue of public safety has attracted much attention and once became a hot topic.With the increase of security awareness,people have gradually realized the importance of identification technology.However,the traditional identification method is simple and easy to forge,which can not meet the needs of the intelligent era.In this situation,biometric identification technology has become the focus of research.Compared with face recognition and fingerprint recognition,gait recognition technology has entered the research area of researchers due to its advantages such as non-contact,long-distance recognition,and difficult to camouflage.At present,gait recognition technology has broad development prospects and research significance in the field of intelligent public safety.However,in practical application scenarios,gait recognition technology is susceptible to many factors.Therefore,in view of some problems in the current gait recognition technology during gait detection and segmentation,and the impact of wearing conditions of pedestrian targets on the gait recognition accuracy,the following researches are performed in this thesis:(1)In the gait preprocessing stage,an object detection and segmentation algorithm based on the optimized ViBe algorithm is proposed.Based on the original ViBe algorithm,the advantages of multiple algorithms are combined to effectively solve the problems of shadows and ghosts during gait detection and segmentation.(2)Aiming at the influence of pedestrian target's wearing conditions on the accuracy of gait recognition,a gait feature extraction algorithm based on gait silhouette contours is proposed.Based on the gait silhouette contours,a segmentation weighting strategy is proposed.The gait silhouette contours are segmented and weighted according to the degree to which the gait feature is affected by the wearing conditions.The gait contour is constructed into a feature matrix to reduce data redundancy.The experimental results show that it can effectively reduce the influence of wearing conditions on the gait recognition accuracy.(3)PCA is used to reduce the gait matrix dimension and reduce the computational complexity.The classification and recognition of gait features were performed using MDA.Classify the targets in the test set by calculating the distance between all the goals in the training set and classify them as the training set targets with which they have the shortest distance.Through a series of comparative experiments on the CASIA B database,the feasibility of the algorithm in this thesis is demonstrated.It is concluded from the experimental results that the gait recognition algorithm proposed in this thesis has a better effect on the shadow problems that occur during gait detection and segmentation and the impact of the wearing conditions of pedestrian targets,and has high robustness and availability.