Research Based On Feature Fusion Of Gait Recognition Under Multi-perspective

With the rapid development of the computer technology and machine visiontechnology, gait recognition is a technology of biometric recognition, whichrecognizes people by the way they walk. Compared with other technologies ofbiometric recognition, gait has its own unique advantages, such as recognizing peoplewithout contact. The gait feature is not easy to disguise，and it can be perceived fromfar away. With the increasing demand for intelligent visual surveillance andmonitoring systems in security-sensitive environment, gait recognition has attracted awide range of research interests.The gait recognition includes moving target detection, feature extraction andclassification and recognition. Based on the study of the various of the gaitrecognition algorithms, this thesis proposes a gait recognition algorithm based on thefusion of shape feature and dynamic feature and makes a detail research on the threeparts.Direct against the limitation of gait recognition method based on only onefeature，a research makes the best of complementary of dynamic information andstatic information and combines them for gait recognition. First step，extracting theimage based on the improved background subtraction algorithm and removing noisebased on morphology，then static information is obtained by analyzing shape of thesubject’s silhouette contours using procrustes mean shape (PMS). After that，theaction energy image (AEI) and the gait energy image (GEI) are calculated. Throughcomparison and analysis，it is concluded that the GEI contains more information onthe dynamic energy. So this test uses GEI which is transformed by Fan-Beamtransform，and then two dimensional principal component analysis (2DPCA) is usedto reduce the dimensions of training and testing data in order to get dynamiccharacteristics of the subject’s frequency. The final recognition result is the fusion of results of two characteristics. The experiments in CASIA gait dataset B gets highrecognition rate and achieves the expected effect of recognition.