Font Size: a A A

The Research On Gait Recognition Method Based On Shape Feature

Posted on:2019-08-10Degree:MasterType:Thesis
Country:ChinaCandidate:Y K LiuFull Text:PDF
GTID:2428330563956441Subject:Public Security Technology
Abstract/Summary:PDF Full Text Request
With the development of social pluralism,people's requirements for security have been continuously improved.The effective identification of human identity under video surveillance is the premise of security protection.Gait recognition is a new type of biometric identification technology.It has the characteristics of non-invasiveness,uniqueness,and easy camouflaging.The gait image acquisition can be completed without the coordination of the movement subject.It is the most effective identification method in the long-distance situation and has become a research hotspot in the field of image processing and identification.Gait recognition methods based on shape features are researched in this paper,including gait detection and image preprocessing,shape-based gait feature extraction,feature fusion and classification recognition,designing and implementing gait recognition software based on shape features which can achieve gait detection and identification.The specific work is as follows:In terms of the gait detection and image preprocessing,gait detection is implemented using a mixture of Gaussian models.Then,morphological,cycle detection,and image size normalization are performed on gait images to facilitate later feature extraction.The simulation experiment results show that compared with the multi-frame average background modeling method,the gait detection integrity of the mixed Gaussian background modeling method is better.In terms of gait feature extraction based on shape,a shape context method based on resampling of feature points is proposed as contour features of gait.The method selects the combination of contour points and joints of lower extremities as sample points.Then the different sampling points are placed at the origin of the polar coordinates,and the number of sampling points located in different sectors is counted as the shape context characteristics.Aiming at the regional features of gait,a calculation method of weighted Hu moment is proposed.According to the change degree of each pixel in the image during the walking,the weighted Hu moments are calculated by giving different weights.Simulation experiments show that the improved methods can describe gait features more effectively and has better real-time performance.In terms of feature fusion and classification recognition,firstly,using a one-to-one voting strategy to design SVM classifiers,the gait recognition based on the feature of shape context and Hu moment is used to realize the gait recognition with a single shape feature,and the effect is better.Then,a multi-feature fusion recognition method is proposed.Gait contour features and region features are integrated in the feature layer and feature vectors are constructed for gait recognition.Simulation experiments show that the fusion feature recognition rate is up to 88.7%,which is higher than that of a single feature.The recognition rate based on the shape context method is up to 85.7%,and the recognition rate based on the Hu moment feature is up to 68.6%.In terms of the design and implementation of gait recognition software based on shape features,the user interface is designed using MATLAB GUI to achieve the detection of moving objects,feature extraction based on shape context and Hu moment,gait recognition,at the same time,the software also can limit user permissions and allow new user registration.
Keywords/Search Tags:gait detection, shape context, Hu moment, SVM classifier, gait recognition
PDF Full Text Request
Related items