Hand veins segmentation and matching under adverse conditions

Abstract

Biometric identification based oil hand veins subcutaneous network structure appears as a promising technique for personal recognition due to its robustness, low cost implementation and high users acceptability. Two of the most critical stages in these vein check identification systems are the vein pattern segmentation and matching, which extremely depend on the image acquisition process: The acquisition becomes a bottleneck in the performance of the whole system, as it represents the first stage in the recognition process. In this paper, solutions for the segmentation and matching under poor illumination. conditions during the image acquisition process and low pixel camera resolution are presented. In particular, exhaustive studies will be presented in order to show that the use of the parametric black tophat transform, instead of the classical tophat mathematical morphology tool, and the utilization of homomorphic filters greatly benefit the performance of the segmentation and matching processes. Furthermore, it will be demonstrated that slight vertical and horizontal hand movements in the image acquisition can be easily corrected by using simple scaling operations. This fact makes unnecessary to fix the hand position and thus, contributes to enhance even more the non-invasive condition of these biometric systems. A first prototype for the image acquisition process has been implemented by only using a very simple CDD camera and LED diodes, obtaining infrared images of 60 people. The application of the mentioned techniques on these images, combined with some typical image processing operations like mean and median filtering, skeletonization and pruning, leads to a reliable vein based identification system, even in poor image acquisition conditions.