Fast exact fingerprint indexing based on Compact Binary Minutia Cylinder Codes
| Autor: | Mingqiang Li, Chaochao Bai, Weiqiang Wang, Tong Zhao |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Minutiae
021110 strategic defence & security studies Cognitive Neuroscience Hash function 0211 other engineering and technologies Binary number 02 engineering and technology computer.software_genre Hash table Computer Science Applications Artificial Intelligence Fingerprint 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Binary code Data mining Hamming space Quantization (image processing) computer Algorithm Mathematics |
| Zdroj: | Neurocomputing. 275:1711-1724 |
| ISSN: | 0925-2312 |
| DOI: | 10.1016/j.neucom.2017.10.027 |
| Popis: | With explosive growth in fingerprint databases, Automatic Fingerprint Identification System has become more challenging than ever. Consequently, it is necessary to develop a fast and exact fingerprint indexing to meet the efficiency and accuracy. In this paper, learning Compact Binary Minutia Cylinder Code (CBMCC) is proposed as an effective and discriminative feature representation and Multi-Index Hashing (MIH) is suitably adopted to accelerate the exact search in fingerprint indexing field for the first time. Firstly, we analyze Minutia Cylinder Code to find that it is strongly bit-correlated and awfully unbalanced. Accordingly, we propose an optimization model to learn CBMCC with the balanced independent property and the minimal binary quantization loss. Finally, MIH method further speeds up the exact search in Hamming space by building multiple hash tables on binary code substrings. The performance test shows that CBMCC is effective and discriminative as it has the maximum intra-bit variance while the minimum inter-bit correlation. Furthermore, numerous experiments on public databases demonstrate that CBMCC–MIH is quite outstanding for fingerprint indexing since it achieves an extremely small error rate with a fairly low penetration rate. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect