Study Of Automatic Eyebrow Recognition Based On Template Matching And Projection Dimension Deduction Analysis

Biometric identification technology is widely used in the field of authentication,because of the characteristics of its security, stability and convenience, etc. Biometricsuch as fingerprint, face, iris, retina, palm shape, signature, and palm prints, etc. iscommonly used in the identification. For human eyebrows could meet therequirements of Universality, uniqueness, collecting, stability, acceptability,performance requirements and security, it could be used for identification studies aswell. The study on eyebrow recognition has great theoretical and practicalsignificance in fields of biometric identification.In this paper, an automatic eyebrow recognition method based on templatematching and projection dimensional deduction analysis was proposed. The mainresearch work as follows:1. To extract pure eyebrow image through the template matching methodusing quasi Haar transformIn this paper, the template matching method using orthogonal Haar transformhave been studied, and on its basis we proposed a algorithm called template matchingusing quasi-Haar transform for non-standard template,using which for pure eyebrowimage extraction. In the algorithm, the quasi Haar features of the candidate window inthe original image and the template are calculated by the tree-decomposition, and thenthe distance between the features was used as filter criteria to locate the best area inthe original image matching with the eyebrow template. Finally, automatic extract thepart of the eyebrows in the face.2. The analysis of the projection dimension reduction method in the pureeyebrow imageThe projection dimension reduction method such as PCA,2DPCA, LPP and2DLPP is studied in this paper. These methods are used to reduce dimension of thepure eyebrow image. Through experiments in the eyebrows database, we compared ofthe difference in its recognition rate, time and dimension.3. Study of automatic eyebrow recognition methodsGiven original eyebrow images and pure eyebrows database, the mission of theautomatic eyebrow recognition is to identify the category of original eyebrows. Theprocess as follow: Firstly, the pure eyebrows in the database was used as eyebrow templates to execute template matching in a given original eyebrow images, theresults will be more than one matching areas; Then, the matching areas were used astest set,the pure eyebrow database as the training set, the eyebrow images of both setsare execute projection dimensional reduction; Finally, according to the distancebetween the projection characteristics, using the nearest neighbor method to get pureeyebrows category of the original eyebrows. During the whole experiment, the pureeyebrow templates in database were selected artificially, the whole identificationprocess is automatic, thereby reducing the artificial interference.