Fusion Approach for Fingerprint Matching for Improved System Accuracy

In automated fingerprint identification systems, an efficient and accurate alignment algorithm in the preprocessing stage plays a crucial role in the performance of the whole system, affecting greatly the speed and accuracy otherwise. This paper proposes a fusion scheme (weighted sum) for aligning the enrolled and query images using modified ring model and cross correlation approaches. Both the methods align the pair of fingerprint images based on the single singular point (as a reference point). Matching is then performed using Euclidean distance based matcher. This model is tested on both publicly available (Cross Match Verifier 300 sensor) as well as proprietary (Lumidigm Venus V100 OEM Module sensor) fingerprint databases scanned at 500 dpi. The experiments show that this fusion approach improves the overall system accuracy: FNMR and FMR significantly dropped to 3.23% and 2.67% respectively for Cross Match Dataset and 0% and 1.33% respectively for Lumidigm Dataset. Hence, the combination of these two alignment methods effectively strengthens the performance of the matcher.