The need for very large fingerprint benchmarks to test the efficacy of emerging commercial technologies for fingerprint minutia detection, matching, compression, and retrieval is well recognized. It is driving more intensive use of networked parallel processing clusters. This report demonstrated the feasibility of a parallel fingerprint matching hosted in the NIST Data Flow System Version II middleware layer, and identified specific hardware and software constraints that cause performance problems when introducing parallelism. The starting point consisted in a standalone matcher that takes two fingerprint images as an input and provides a matching score by using the well-known Bozorth fingerprint matcher. The publicly available NIST Special Database 14 and tools were used for matching and in order to handle fingerprint images and perform the minutia detection and print matching. Scalability to about 100 compute cores was shown using conventional gigabit Ethernet connectivity.
Citation: NIST Interagency/Internal Report (NISTIR) - 7754
NIST Pub Series: NIST Interagency/Internal Report (NISTIR)
Pub Type: NIST Pubs
Fingerprint Matching, Parallel Processing