Palm vein recognition using scale invariant feature transform with RANSAC mismatching removal

Palm vein recognition has been gaining increasing interest as a biometric method, although there still remains an issue regarding difficulties in obtaining robust signals. In this paper, the effects of random sample consensus point mismatching removal and the use of different wavelengths of illumination on the recognition rate are investigated. The CASIA multi-spectral palm print image database was used to provide input signals and the scale invariant feature transform (SIFT) and random sample consensus (RANSAC) mismatching removal approaches were adopted for vein extraction and point feature matching. The results show that the RANSAC mismatching point removal was able to eliminate outliers while preserving the appropriate SIFT key points and that this led to an improvement in the equal error rate metric, signifying better recognition performance. The palm vein recognition system was found to achieve a better verification rate when infrared illumination in a specific spectral band was used to obtain the palm vein image.