Font Size: a A A

Research On Gait-based Remote Human Recognition System

Posted on:2014-02-10Degree:MasterType:Thesis
Country:ChinaCandidate:H Y ChenFull Text:PDF
GTID:2298330422968259Subject:Instrument Science and Technology
Abstract/Summary:PDF Full Text Request
Along with the increasing demand on the intelligent monitoring at military area,sensitive public occasions, and national important department, remote humanrecognition has been becoming a hot task, and the traditional identification cannotqualified this task. Traditional biometrics (such as fingerprint, face, and iris) cannotcomplete the task of remote identification. Gait recognition is a new biologicalrecognition technology, in order to extract their respective characteristics formdifferent people’s behavior of walk, it can achieve the goal of automatic identification.The gait feature has some advantages that it is difficult to disguise and hidden, notcontact, distance sensing, and can be detection and recognition at the low resolution inimage sequence. Therefore, from the perspective of safety monitoring, gait is the mostpotential biological characteristics at the distance cases. The gait recognitiontechnology has very wide useful range at security sensitive sites and special personnelrestrictions.Based on the existing algorithm and theory, this article emphatically from the gaitcontour’s Fourier descriptor, human body articulation’s angle, reflex symmetry factor,we conduct research at those aspects, the main work is as follows:1、First, the paper puts forward the gait recognition system overall design goaland the implementation plan, we have applied the CASIA gait database, the paperhave selected the Data Set B as the experiment database, we choose some videosequence that from the Data Set B as the experimental samples, to verify thefeasibility of the experiment scheme and recognition effect. According to thelaboratory’s existing laboratory equipment, shot20video clips at inside and outsidethe laboratory, those video will be used in comparative.2、After the motion object is detected, the image will be cut and reduced, usingthe double liner interpolation method to reduce the image, this can reduce the processamount of the image data.3、According to the cyclical changes of the people’s step length to measure thegait cycle, at one gait cycle, we extract the target’s three characteristics: body contourfeature, limb anger feature, and the body ‘s reflection symmetry feature. In order toreduce the dimension of feature, take the first30step Fourier transform of profile distance vector, take nine angle value of human body, and take one reflectionsymmetry factor, those data combine40dimension characteristic vector, used forclassification and recognition.4、In order to verify the efficiency of the method, the experiment part is dividedinto four groups, the first group use two features for identification, the second groupcombine three features, but without introducing weights, the third group use theweight, the fourth group use our own shoot video for test.5、We completed the preparation of the software system independently, comparedthe four above algorithm results, the recognition algorithm proposed in this paper canachieve a recognition rate over90%in CASIA gait database; in the video shotourselves, the recognition rate can reach85%.
Keywords/Search Tags:gait recognition, reflective symmetry factor, Fourier descriptor, feature fusion, nearest neighbor fuzzy classifier
PDF Full Text Request
Related items