Font Size: a A A

Gait Recognition Research Based On The Spatial Distribution Of Optical Flow

Posted on:2013-03-21Degree:MasterType:Thesis
Country:ChinaCandidate:Y YangFull Text:PDF
GTID:2268330392970173Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
Biometrics makes use of the physiological or behavioral characteristics of peopleto authenticate their identities. With the growing need for visual surveillance andmonitoring system in security-sensitive environment such as airports, bus stops andbanks, human identification at a distance has recently gained increasing interest fromresearchers. The established biometrics such as face, fingerprints and iris usuallyrequire proximal sensing or physical contact. However, they are hardly applicable at adistance. Fortunately, gait is still visible and can be easily perceived unobtrusively. Sogait is a very attractive modality from the visual surveillance perspective.Silhouette extraction is an important procedure in gait recognition, whichincludes two steps: background modeling and motion extraction. Kernel GaussianModel is used to model the gait background, and kernel width estimation method isproposed. After gait images are mapped into probability images by the statisticalbackground model, foreground pixels are extracted from these probability images bythe Expectation-maximization algorithm. A novel gait recognition method based onspatiotemporal feature extraction is proposed under the situation of bad silhouettequality. This paper proposed the spatial distribution of the optical flow to descript andrecognize moving target. First, compute dense optical flow for each image in asequence and derive scale-independent scalar features which characterized the spatialdistribution of the flow. Then analyze periodic structure of these sequences of scalars.The scalar sequences for an image sequence has the same fundamental period butdiffer in phase. The phase feature vectors can be used to recognize individuals. Lastly,for each sample, train the average of feature vectors as cluster centers. Sequenceswere classified to the nearest class based on the nearest neighbor rule.The experiment results show that, the proposed method achieves obviousimprovement than traditional methods for the complicated scenes. And at the meantime´╝îit can satisfy the requirement for real-time background modeling. The proposedmethod achieves highly competitive performance and it takes less storage andcomputational cost.
Keywords/Search Tags:Biometrics, Gait Recognition, Feature Extraction, Classificationand Identification
PDF Full Text Request
Related items