Research On Key Techniques Of Image Calibration For Contactless Palmprint Recognition

With the arrival of the age of smart city and intelligent home, people nowadays have a higher requirement for security and usability of authentication based on biometric features. At the same time, security situation continues to deteriorate, and biometrics is now urgently needed in security field. Palmprint recognition technology can meet those demands since the palmprint texture is feature-rich and stable. And it is quite a nature way for users to make authentication just by using their hands. Hence contactless palmprint recognition is now rapidly developed. For contactless palmprint system, users do not need to let their hands be posted on the device. Thus it provides a fresh way which is free and hygienic to use. But the rejection rate of contactless palmprint recognition system increases because of the complex capture environment and image distortions.This paper aims to solve those typical problems which exist in real environment,and corresponding calibration methods for images which suffer distortions from rotation, translation, scale change, blur and tilt will be researched on. Our work includes the following points:First, Image calibration algorithm is realized to correct images which are imaging from a tilt palm plane that has an angle between the ideal plane. And at last experiments are designed to test the accuracy of the tilt calibration system.Second, A robust palmprint center block localization algorithm based on SIFT is presented. It can correct the in-planar distortion such like rotation, shift and scale change. The rotation distortion is in the range of 60 degrees.Third, the number of SIFT matching points is significantly increased by different image enhancement methods and scale space parameter optimization methods. At last the traditional ROI localization algorithm is improved and based on this localization algorithm a SIFT-Comp Code joint decision-making palmprint recognition system is realized.In this paper, we implement an image calibration module for the contactless palmprint recognition system, which detects distortions of the current input image,and the corresponding calibration operation is processed. It improves user experience and guarantees accuracy of the palmprint identification system at the same time.