研究目的
To develop a low-cost multi-fingervein verification system that captures three different fingers simultaneously to improve accuracy and reliability in biometric applications.
研究成果
The proposed multi-fingervein sensor successfully captures three fingers simultaneously, improving verification accuracy and reliability. Score level fusion enhances performance, with low EER values, making it suitable for applications like access control. Future work could expand the database and test in more diverse conditions.
研究不足
The study is limited to 20 subjects and three specific fingers (index, middle, ring), excluding thumb and little finger due to design challenges. The sensor may have issues with uniform illumination and finger placement errors, potentially affecting image quality. The database size is small, and real-world deployment scenarios might require further testing for robustness against environmental variations and presentation attacks.
1：Experimental Design and Method Selection:
The experiment involves designing a multi-fingervein sensor using near-infrared (NIR) light penetration method with a single camera and specific illumination structure. Four state-of-the-art feature extraction algorithms (MCP, SMR, RLT, WLD) and a probabilistic collaborative representation classifier (P-CRC) are used for verification.
2：Sample Selection and Data Sources:
A new database is created with 20 unique identities (subjects), capturing three fingers (index, middle, ring) from each hand in 10 sessions, resulting in 600 fingervein images.
3：List of Experimental Equipment and Materials:
The sensor includes NIR LEDs (TSFF5210 from Vishay semiconductors), a camera (DMK 22BUCO3 from Imaging Source), a lens (focal length
4：5-8mm from Computar), and a physical structure for light scattering and finger placement. Experimental Procedures and Operational Workflow:
Fingers are placed in slots to restrict rotation; NIR light at 870nm wavelength illuminates the fingers, and images are captured. Image enhancement is performed using adaptive histogram equalization, followed by feature extraction and comparison using P-CRC. Score level fusion (SUM rule) is applied to combine results from three fingers.
5：Data Analysis Methods:
Evaluation uses leave-one-out cross-validation protocol, with performance measured by Equal Error Rate (EER) and analysis of genuine and impostor score distributions.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
