Research And Implementation Of Knuckleprint Recognition System And Key Technologies

The arrival of the information age is improving the status of information security. Being an important method of authentication, biometric identification technology always is the research mainstream. As using the human inherent physiological and behavioral characteristics, biometric identification technology shows great advantage than any other technologies. Traditionally, biometric identification technology has some shortcomings, consider of this situation, this paper brings forward knuckleprint as a new biometric identification technology to overcome the shortcoming and exploit another biometric identification method.First, this paper analyzes the characteristics of knuckleprint, principles identification system, functional requirement and capability guide line and design scheme. Identification of key technologies knuckleprint also carried out a detailed summary.Second, in order to solve the problem of accurate location of finger, this paper utilizes a smallest eigenvalue analysis method to locate the finger fiducial point. Based on uncontinuity of curve, this method uses single pixel and a curve which contain the single pixel as analysis object, and then utilizes macroscopical collinearity to judge the target curve is the region of finger fiducial point or not. Owing to this, the smallest eigenvalue of covariance matrix could get from these target pixels. Finally, through the critical point analyzing, the finger fiducial point is located exactly. Emulation experiment shows that this algorithm overcomes disadvantages in traditional method such as time-consuming, complicated and inaccurate. At the same time, a finger segmentation technology based on straight-line fitting is adopted in this paper. This technology utilizes the linetype characteristic of finger contour to realize finger segmentation in circle.Third, in order to solve the problem of accurate location of region of interest of knuckleprint, by giving the definition of ROI of knuckleprint, this paper brings forward a new automatic detection and location algorithm based on wavelet multi-resolution to locate the ROI of knuckleprint. Based on texture similar theory, this algorithm uses feature vector and regional growth to produce candidate sub-region set in high frequency sub-images, and then utilize Radon projection in low frequency sub-images to verify the candidate sub-region set. Finally, by adopting straight-line fitting technique, the location of ROI of knuckleprint in original image is accurately located. At the same time, a new segmentation method of knuckleprint based on modified Gabor filter is utilized in this paper. This method uses finger axis direction and certain frequencies as Gabor filter's parameters. Emulation experiment shows that this algorithm is more suitable for knuckleprint.Fourth, a new matching algorithm of knuckleprint based on moment invariants and wavelet multi-resolution analysis is presented in this paper. In the first recognition stage, in order to overcome the defects caused by scale factor and rotation angle to get the set of candidate images, the algorithm not only proposes an idea of minimal rectangle to modify the calculation region of moment invariants but also adopts the elementary row transformation to transform the character matrix of radon projection. In the second recognition stage, the wavelet multi-resolution analysis is adopted to analyze the certain angle projection of candidate images set. Finally, the weighted character of moment invariants is calculated to get the similarity. Emulation experiment shows that this algorithm is more efficient than traditional algorithm. At the same time, a new fusion technology of weighted information from knuckleprint and palmprint is proposed in this paper. First of all, this algorithm proposes a self-adaptive method to locate ROI of palmprint and complete segmentation by double stages wavelet analysis, and then, fuzzy theory is adopted to finish fusion matching.Finally, this paper describes software design in frame, algorithm parameters and processing results. Two sample databases are utilized to verify the robust of this system, and the results show that this system satisfies requirements of biometric identification system.