Research On Identity Recognition Technology Based On Gait Features

With the national emphasis on security,video surveillance in public places has become more and more important,and it has important research significance to identify the pedestrian identity through the monitoring system.Because the camera is generally far away from the pedestrian,the captured image is not clear and the face is blurred.Traditional methods for biometric identification,such as Face Recognition and Iris Recognition,have been unable to meet the actual demand.Gait recognition,because of its advantages of long-distance recognition and low resolution,has obvious advantages over traditional biometric methods,and has become a research hotspot in the field of computer vision.At present,gait recognition is mainly divided into two categories:model-based approach and non-model based approach.The gait energy map belongs to the non-model-based method.Because the gait energy map contains richer dynamic and static information,compared with a gait recognition method based on the model has the advantages of easy extraction,feature rich,so this article mainly studies the gait recognition based on gait energy diagram.Identification method based on gait energy diagram includes gait preprocessing,feature extraction,gait recognition three steps.How to extract features effectively is the key to gait recognition,and it is also the key research content of this article.In this article,the main research work and contributions are as follows:(1)Based on gait energy diagram,an improved method to extract gait energy figure-frame difference threshold method,the gait energy map extracted by this method can reduce the influence of the static information on gait recognition while retaining the high frequency walking information of the hand and leg.(2)To reduce the recognition rate caused by interference factors such as knapsacks and overcoats,put forward a kind of weighted segmentation algorithm,the algorithm can well highlight the contribution of each part of the energy map,thereby reducing the impact of backpack,coat and other interference factors on gait recognition,and achieving the goal of improving the recognition rate.(3)Using the five-scale and eight-angle Gabor wavelet to characterize the segmented weighted energy map,the weighted filter features are obtained,and the effectiveness of the extracted features is proved through comparative experiments.(4)In view of the Gabor wavelet feature extraction,high dimension problems,using 2DPCA method to feature dimension reduction of data,then using the SVM classifier to identify the gait features with strong generalization ability.The experiment shows that the recognition method based on 2DPCA+SVM can improve operation efficiency and achieve good recognition effect.