Research On Character Extraction And Recognition Method Of Hand Shape

As the fast development of economy and society, the information security and social safety has becomes a key problem which raises people's more and more attention and higher demanding of identity recognition. Identity recognition has grown to an crucial means to ensure the security of information and network since traditional security technology can not satisfy the higher demanding of security due to its intrinsic defects such as card can be easily fabricated and password can be stolen. Compared with the traditional methods, recognition of person's physical features has become the main means since they are unique and invariant.As one of the main biometric recognition technology, hand-shape recognition is another principal method of biometric recognition beyond the fingerprint recognition due to its simple sampling device (compared with fingerprint, iris etc.) and fast recognition speed. Currently, in the biometric product markets aboard, products of hand-shape recognition gets the same rate with that of fingerprint recognition and it has becomes another large category recognition products. This dissertation adapts grey scale hand-shape images, takes advantage of the image processing and pattern recognition technology to automatically realize a series of procedures such as the pre-processing, feature extraction, pattern matching and so on, successfully designs and achieves a hand-shape recognition/verification system and get a good experimental results.This dissertation takes research in the kind of hand picture which is captured in an non-touch mode and the fingers stretching out having no touch with each other and there are no tails to fix the hand. To solve the problem that it is difficult to extract the accurate hand contour when hand images are affected by the unsymmetrical illumination, this dissertation proposes a contour tracking algorithm based on the directional maximal gradient value. This algorithm can directly track the accurate, consecutive and integral hand contour in gray-level picture and it especially adjusts to the contour extraction of hand images affected by the unsymmetrical illumination. The dissertation also presents a method for personal identification based on hand geometry feature according to the characteristic that hand shape is measurable. The method uses finger length,finger width and palm width etc. to compose relative length eigenvector, and achieve personal identification by computing the distance between eigenvectors. It can reach a high recognition rate as over 90% only use a few characters, and this method is easy to realize,only need to store a few characters and is suited to realize in hardware.