![]() | Only 14 pages are availabe for public view |
Abstract The need to identify people is as old as humankind. People recognize each other by sight and sound. However, in today ’s complex society, it’s impossible to personally know everyone. Biometric devices automate the personal recognition process. Each of us is unique from every other human being. We have unique physical characteristics, such as hand shape, blood vessel patterns and fingerprints. Biometric devices measure and record these characteristics for automated comparison and verification. The thesis presents an Automatic Hand Geometry Verification System (AHVS) which is comprised of: (1) Data acquisition stage. (2) Preprocessing stage where problems such as: i-Binarization, ii-Edge detection, iii-Thinning and iv-Tracing are covered and solutions are provided. (3) Feature extraction stage where the feature vector is obtained. (4) Matching stage (which is the major contribution of this thesis) with a reference database of hand images. A new algorithm is proposed for the extraction of the hand geometry characteristic features based on anatomical landmarks (fingertips and bottom of valleys between fingers). These features are extracted from the plan of the 2D-hand image captured by an infrared CCD camera. These features can be used for the verification process. Image processing operations are used to get forty five anatomical landmarks. Thirty measures are extracted automatically based on this 45-points hand model. These features are the finger lengths, finger areas, finger circumferences, and finger widths at different heights of the different fingers of the right hand as well as the radius of the hand center. At last, in order to evaluate the matching stage, the algorithm has been tested on an experimental data set collected by the designed system consisting of 500 hand images for 100 persons from the general public in different ages, and it showed up to 99.68 percent rate of success. |