HAND GEOMETRY BASED HUMAN IDENTIFICATION

Abstract

Hand geometry recognition systems are based on a number of measurements taken from the human hand, including its shape, size of the palm, and the lengths and widths of the fingers. The protection of personal information against criminals has become very important and with biometric recognition systems finding applications in many security information systems with varying requirements, the use of biometrics as a means of accessing information has proven to be away to better these incidents. Hand Recognition is dependable for these objectives. Hand recognition process consists of the image acquisition, preprocessing, detecting the peaks and valleys of the hand and features extraction when compared (matching) with the database. Hand geometry recognition requires an efficient algorithm to accurately detect peaks and valleys and the robustness of the system. Several existing algorithms have been developed which are only useful for detecting the peaks and valleys of either the right or left hand but not both. This research presents a new algorithm for peaks and valleys detection and Manhattan model for recognition for both hands. A methodology was developed based on Canny Edge/Manhattan. Once the points have been located, the necessary features for authentication were extracted prior to the implementation of the classification algorithms. This process was based on the detection of five points (peaks) that correspond to the fingertips and the four points between them (valley). Specific methods often have to be implemented during the acquisitions stage to make the detection of these points easier. The acquired handimages were preprocessed using the canny edge detection algorithm while the developed algorithm was used to detect the peaks and valleys. 5 peaks and 4 valleys were detected which generated 31 distances measured as the extracted features and were stored as biometric templates. The Manhattan distance was employed for the classification of the template and recognition. The hand geometry verification system was tested by using a database of 300 images. Database of this system consists of 10 different acquisitions of 30 people. Most of the considered users were within a selective age range from21 to 30 years old though the system works for all adult age brackets. Six images of each user’s hand were selected to compute the feature vector which is stored in the database along with the user’s name. The remaining four images were used for verification. The verification scheme depends on the features vector. Verification refers to the process of confirming or denying a claim of an individual. This is known as one-to-one matching. The proposed system achieved an accuracy of 97% with an FAR of 2% and FRR of 3% was achieved.