Research On Character Extraction And Recognition Method Of Deformable Hand Shape Image Contour

In today’s information age, the information security and social safety has becomes akey problem which raises people’s more and more attention and higher demanding ofidentity recognition. Identity recognition has grown to a crucial means to ensure thesecurity of information and society. As a kind of biometric technology, recognition ofperson’s certain physical features has gradually become the main means and method sincethey are unique and invariant, which is the main development direction for the future ofsecurity technology.As one of the biometric recognition technology, hand shape recognition is a majoridentification methods following the fingerprint identification. It has the advantages of lowcost, simple operation and quick certification, so it has been widely used in access control,time and attendance system. Currently, as another major categories of biometric products,the hand shape recognition products have been basically flat with the fingerprintidentification products at the biometric market share of goods in foreign countries. Thisdissertation adapts grey scale hand-shape images in BMP formation, takes advantage of theimage processing and pattern recognition technology to automatically realize the a series ofprocedures such as the pre-processing, feature extraction, pattern matching and so on.This dissertation takes research in the kind of deformable hand shape picture which iscaptured in an contact mode and there are no tails to fix the hand, Allowing the hand has acertain extent the direction of change.It is called rigid body deformation .First, to solve theproblem that it is difficult to extract the accurate hand contour when hand images areaffected by the unsymmetrical illumination, this dissertation proposes a contour trackingalgorithm based on the directional maximal gradient value. This algorithm can directlytrack the accurate, consecutive and integral hand contour in gray-level picture and itespecially adjusts to the contour extraction of hand images affected by the unsymmetricalillumination. Second, the dissertation also presents a method for personal identificationbased on hand geometry feature according to the characteristic that hand shape ismeasurable. The method uses finger length and finger width to compose relative length eigenvector, and achieve personal identification by computing the distance betweeneigenvectors. It can reach a high recognition rate as choose 6 characters. Finally, throughthe establishment of different deformation degree of hand-shaped gallery, the identificationof rigid deformed hand is realized by the application of the hand in different parts of thegeometric characteristics. The recognition rate is 74%.