Document Type



Australian Institute of Security and Applied Technology, Edith Cowan University

Place of Publication

Perth, Western Australia


Australian Institute of Security and Applied Technology


Mehnert, A. J., Cross, J.M., Smith, C.L., & Chia, K.Y. (1995), A personal identification biometric system based on back-of-hand vein patterns. Perth, Australia: Australian Institute of Security and Applied Technology, Edith Cowan University.


This report describes research on the use of back-of-hand vein patterns as a means of uniquely identifying people. In particular it describes a prototype biometric system developed by the Australian Institute of Security and Applied Technology (AISAT). This system comprises an infrared cold source, a monochrome CCD camera, a monochrome frame-grabber, a personal computer, and custom image acquisition, processing, registration, and matching software. The image processing algorithms are based on Mathematical Morphology. Registration is performed using rotation and translation with respect to the centroid of the two-dimensional domain of a hand. Vein patterns are stored as medial axis representations. Matching involves comparing a given medial axis pattern against a library of patterns using constrained sequential correlation. The matching is two-fold: a newly acquired signature is matched against a dilated library signature, and then the library signature is matched against the dilated acquired signature; this is necessary because of the positional noise exhibited by the back-of-hand veins. The results of a cross-matching experiment for a sample of 20 adults and more than 100 hand images is detailed. In addition preliminary estimates of the false acceptance rate (FAR) and false rejection rate (FRR) for the prototype system are given. Fuzzy relaxation on an association graph is discussed as an alternative to sequential correlation for the matching of vein signatures. An example is provided (including a C program) illustrating the matching process for a pair of signatures obtained from the same hand. The example demonstrates the ability of the fuzzy relaxation method to deal with segmentation errors.