The Research Of Method In Gait Recognition Based On Video

With the development of society,people pay more attention to the security issues seriously.Recently biometric identification has gradually become a focus in the field of identity recognition.Gait recognition is one of the new identification technologies.It intends to base on the movement of people's walking achieving the identification.It is also considered to be an important recognition method in a long distance.Gait recognition includes the technologies of digital image processing,computer vision and pattern recognition and it is also considered the most potential recognition method in the future.There are 3 parts which include detection of gait objects,features extraction and gait recognition in this thesis.The main research includes:Firstly,a new algorithm of feature extraction,the Gait Energy Image with Points(PGEI)is proposed in this thesis.Pre-processed gait silhouettes are cut,and then trace the contour of the cut images.The key points of human body in the contour image are positioned,such as knee joints,ankle joints.Then the gait characteristics are expressed completely through PGEI which is formed with the thought of GEI.This method not only can express the characteristics on the gait contour,but also display the trajectory of key points in the PGEI.And more gait features of a gait cycle are concentrated in the image.After getting the map of PGEI,the texture features are extracted by using algorithm of Local Binary Partten(LBP)which has advantages of high computing efficiency and good robustness.Secondly,a method of inter Frame Gait Energy Image(FGEI)is proposed to express totally gait features.Each frame has coupling with the adjacent frames in the gait silhouettes.In order to express the characteristic between images,the frame difference images between the current frame and the previous or next frame silhouette are made.And then the two difference maps are superposed to obtain the inter frame gait energy image.In a gait cycle,4 frames of the maximum and minimum distance between two legs are chosen as the key frames for features extraction,and the row mass vector and column mass vector are extracted for the next recognition.This method can express the gait characteristics with the movement trend and the coordination between limbs and trunks completely in two dimensions.At last,a gait recognition method based on multi features fusion with support vector machine is researched.The row mass vector and column mass vector that extract from the key frame maps are fused.According to the classification error rate of each feature,a judgment formula is established with calculate weight.Then SVM is used for gait recognition.This approach covers more gait features and has better recognition result than single feature.The accuracy rate is 94.44%,which higher than single row mass vector accuracy rate 11.11%.The rate is also higher than single column mass vector rate 16.66%.