Research On Fusion Recognition Of Forearm Contour/Vein Biometrics

With the rapid development of information technology,information security,especially personal identity information security becomes more and more important.In daily life,personal identification is widely used in public security,judicial and other various fields.The traditional identity authentication methods have been unable to meet the needs of modern society because of their own defects.Biometric identification has the advantages of security,reliability and convenience,and gradually replaces the traditional identification technologies.At present,most biometric identification techniques are based on a single mode,which has some deficiencies in the practical application due to the data noise and easy forgery.Multi-modal biometrics give full play to the advantages of different biometric features,reduce the adverse effects of each biological feature,and further improve the performance of recognition system.Thus,it can overcome some shortcomings of single mode recognition to some extent.This paper mainly focuses on the multi-modal biometrics based on hand shape and forearm veins.After studying the single mode recognition based on hand shape and vein.Then,it carries on the dual mode fusion recognition at the score level.The main work of this paper can be summarized as the following three aspects:1.Feature extraction and recognition based on hand contour: The shape of hand contour is easy to produce non-rigid deformation.This paper uses the geometrical feature of the hand for human identification.Firstly,the complete hand contour is extracted according to the near-infrared hand image.Since other method cannot precisely locate the feature point,the approximate curvature method is used to locate the finger valley and fingertip point and so on.Then the feature points are used to further locate other feature points.All the geometrical features of hand are calculated by the feature points.So,the relative geometrical features are used to characterize the hand shape.At last the hand shape is matched according to the similarity of the relative feature vector of the hand shape.An ideal identification result is obtained.2.Feature extraction and matching based on arm veins: The CLAHE pretreatment is used to enhance the contrast between the skin and vein in the near infrared images.Gabor filter is chosen for vein image processing for its good performance in extracting target.The veins are positioned and enhanced according to the Gabor orientation and energy maps.Then we use the hat and the morphological methods to extract the venous lines and to trim the burrs.Then we sample the venous lines by points.The coherent point drift algorithm(CPD)can better solve the problem of non-rigid deformation matching,which is chosen to identify the point pattern.The high accuracy of venous recognition has obtained.3.Hand shape and vein dual-mode score-level fusion recognition: Due to the shortcomings of sensor-level,feature-level and decision-level fusion,the two biometrics of hand contour and arm vein are fused at the score-level.Firstly,the fixed weight fraction and the individual weight scores are calculated based on CMC Curve and Roc Curve.The max-min rule is used for the matching scores normalization.Then the weighted fusion is carried out.Finally,the hand shape and vein biometric database are used for the fusion experiment.The more stable fusion effect is obtained.The validity and robustness of the algorithm are verified.The reason why the “weight based on area” method is better than the “weight of FIR” method has been proved.Compared with the ERRW and NCW weight calculation methods,the algorithm we used is superior.