Research On Palmprint Recognition System Based On Principle Line Features

Nowadays, people attach more importance to technology security with the rapid growth of information technology. Biometrics is the most secure and convenient way to satisfy the requirements for identity digitalization and virtualization in the coming network society, which combines biology technology and information technology to exploit physical features in human body such as fingerprint and palmprint or behavioral features such as gait and signature to identify a person. Many biometric systems have been developed for various commercial applications such as banking, airport security control and access control. Compared to other biometric technologies, palmprint is in its infancy. But palmprint has several advantages: palmprints contain many distinctive features such as principal lines and wrinkles, which can be extracted from low-resolution images; palmprint capture devices are comparatively cheap, a highly accurate biometrics system can be built by combining all features of palms, such as palm geometry, ridge and valley features, and principal lines and wrinkles, etc. Therefore palmprint recognition has recently attracted an increasing amount of attention from researchers.There are many features in a palmprint such as geometrical features, principal lines, wrinkles, delta points, minutiae, etc. However, geometrical features, such as the width of the palm, can be faked easily by making a model of a hand. Delta points and minutiae only can be extracted from the fine-resolution images. Line features are very important to discriminate between different palmprints and they can be extracted from low-resolution images. So line features are one of the most important features in automated palmprint recognition.In this paper, we concentrate on developing a palmprint recognition system base on principal line features. Based on the palmprint images captured by digital cameral, we propose a novel approach for palmprint alignment and segmentation. Then, focus on the research of line feature extract, and present a new method to extract principal lines simply and effectively. Finally, we design coarse to fine level recognition algorithm based on the extracted features to complete palmprint recognition, and develop recognition software to test the validity of the algorithm.We focused on the research of the key technologies according to the characteristic of palmprint. And propose an efficient algorithm of palmprint image alignment and segmentation, line feature extraction, match and recognition, etc. To be more specific, the main achievements in this paper are as follows:1. Palmprint alignment and segmentationIn this system, no guidance pegs are fixed in our image capture device. Thus, palmprint images with different shifts and rotations are produced. Therefore, in order to decrease the distortion and error, we should define a coordinate system that is used to align different palmprint images for matching. The gaps between the fingers are utilized as reference points, and then a smaller region from the center of the palm, called region of interest (ROI), is automatically extracted.The ROI is defined in square shape and it contains sufficient information to represent the palmprint for further processing.Previous research on palmprint identification mainly focuses on feature extraction and representation. But a crucial issue, palmprint alignment, is not usually addressed. In this paper, a new algorithm for palmprint image alignment and segmentation is proposed. In order to align palmprints, palmprint contour point corner is used as basic point to establish coordinates and obtain the largest center area of palmprint. First, we utilized background segmentation algorithm which can avoid error and uncertainty from setting thresholds to obtain binary image directly. Then radical template is adopted to detect corners. The method is faster, less computationally intensive and has better detecting and alignment performance. Then categorize these potential points according to width of finger, the point which has maximal target area of the template will be the optimum point. The experiments show that this method has low computational complexity, high accuracy.With this method, we can get enough information from palmprint; in another aspect, as the sizes are different between two persons, the length of ROI can be selected as a feature for palmprint matching.2. Palm-line extracting and construction of feature vectorPalm-lines, including the principal lines and wrinkles, which is one of most important features of palmprint, can describe a palmprint clearly. They have stability and uniqueness, and can be extracted from low resolution image. Stable,simple line features of palmprint are the key issues for palmprint matching. Because of the noise and wrinkles it is pretty difficult to extract principle lines from palmprint images. So how to extract principle line features of palmprint from low resolution image is the key issue in palmprint recognition system.This paper analyzed palmprint image, proposed a novel approach of palm-line extraction using diversity and contrast based on their characteristics of ridge edges. Then, improved Hilditch algorithm is used. An approach for getting rid of twig which brought forward by this dissertation is applied, and then the broken line was connected; finally, palm-line image of single pixel is obtained. By this operation, most of principle line can be detected correctly, the experiment result is encouraging.It is difficult to express palm-line accurately. In this paper, palm-line image is divided into 9 overlapped small blocks, where each block overlaps 8×l /3(l equal to the length of ROI) pixels of the each adjacent block and the endpoints of the line segment is used to represent a line. Then line feature is constructed based on the characteristics of palm-lines. The feature vector of palmprint is obtained by combining the length of ROI and principle line feature.3.Design a matching algorithm based on two-level strategyFeature matching is one of the key steps in palmprint identification. Palmprint identification is one-to-many matching process, which relies on the amount and organization mode of data. Since we have to compare with all data of a huge database, the feature extraction and matching algorithms can not be too complicated. Actually, the accuracy and efficiency of recognition algorithm are mutually restricted. Two steps method is usually used to solve the problem. And the method can solve the contradiction of accuracy and efficiency in a large extent.In this paper, a two-stage method of geometrical feature matching and line feature matching is proposed. According to the length of ROI, we can extract the classes that have similar feature with the sample image. Then we start accurate matching using the line feature.A palmprint recognition system is developed with functions of ROI alignment, features extracting, user enrollment and user recognition. The performance of the system is tested on our database. The experimental result indicate that the system achieves high recognition rate, proved that the method is effective and available.