• Không có kết quả nào được tìm thấy

OER000001437-1.pdf

N/A
N/A
Nguyễn Gia Hào

Academic year: 2023

Chia sẻ "OER000001437-1.pdf"

Copied!
268
0
0

Loading.... (view fulltext now)

Văn bản

The first part provides a description of the state of the art in vascular biometry, including an extensive bibliography. The first part provides a description of the state of the art in vascular biometry, including an extensive bibliography on the subject.

Introduction

Hand and Finger Vein Biometrics

Sclera and Retina Biometrics

Security and Privacy in Vascular Biometrics

Chapter 15, by Vedrana Krivokuća and Sébastien Marcel, On the Recognition Performance of BioHash-Protected Finger Vein Templates, applies BioHashing to finger vein templates generated by classical binarization feature extraction and evaluates the resulting recognition performance. Chapter 16, by Simon Kirchgasser, Christof Kauba, and Andreas Uhl,Cancellable Biometrics for Finger Vein Recognition—Application in the Feature Domain, applies the classical cancelable transformations, i.e.

Introduction

State of the Art in Vascular Biometrics

Introduction

  • Imaging Hand-Based Vascular Biometric Traits
  • Imaging Eye-Based Vascular Biometric Traits
  • Pros and Cons of Vascular Biometric Traits

Finger vein recognition deals with the vascular pattern inside the human fingers (this is the latest feature of this class and is often [126] believed to be its origin), while hand/palm/wrist vein recognition visualizes and acquires the pattern of the vessels in the hand central area (or wrist area). With transillumination, the light source and the camera are on the opposite side of the hand, i.e.

Commercial Sensors and Systems .1 Hand-Based Vascular Traits

  • Eye-Based Vascular Traits

The most recent scanner is the ICAM 2001 model, and it appears that this device is still for sale.3 In the first decade of this century, the company went. For sclera biometrics, the startup EyeVerify (founded in 2012) called their sclera recognition technology "Eyeprint ID", for which the company also acquired the associated patent.

Algorithms in the Recognition Toolchain

  • Finger Vein Recognition Toolchain
  • Palm Vein Recognition Toolchain
  • Wrist Vein Recognition Toolchain
  • Retina Recognition Toolchain
  • Sclera Recognition Toolchain

Classical techniques that result in a binary layout of the vascular network (which is usually used as a template and is subject to correlation-based template comparison using alignment compensation) include iterative line tracking [174], maximum curvature [175], principal curvature [ 32], mean curvature [244] and broad line detection [94] (where the latter technique suggests a comparative degree of finger rotation compensation template). Finally, [9] rely on their approach of using biometric graph matching (BGM) on graphs obtained from vascular network skeletons.

Table 1.1 Finger vein feature extraction techniques focussing on vascular structure
Table 1.1 Finger vein feature extraction techniques focussing on vascular structure

Datasets, Competitions and Open-Source Software .1 Hand-Based Vascular Traits

  • Eye-Based Vascular Traits

As far as the authors are aware, no public competition has been announced in this field. In the context of the (medical) analysis of the retinal vasculature, software was provided for retinal vessel extraction based on wavelet domain techniques: the ARIA Matlab package based on [12] and another MATLAB software package called mlvessel8 based on the methods described in [ 241].

Table 1.5 Finger vein datasets available for research (typically upon written request) Name Dors/palm Subjects Fingers Images Sess
Table 1.5 Finger vein datasets available for research (typically upon written request) Name Dors/palm Subjects Fingers Images Sess

Template Protection

  • Hand-Based Vascular Traits
  • Eye-Based Vascular Traits

In order to account for the weaknesses revealed in the soft vault scheme due to non-uniformity in biometric data, two-factor authentication using an additional BCS hardening password is proposed. In the context of sclera recognition, [189] proposes a CB scheme based on a region indicator matrix that is generated using an angular grid reference frame.

Presentation Attacks and Detection, and Sample Quality

  • Presentation Attack Detection
  • Biometric Sample Quality—Hand-Based Vascular Traits
  • Biometric Sample Quality—Eye-Based Vascular Traits

The authors of [251] propose windowed dynamic decomposition (W-DMD) to be used to identify fake finger vein images. Finger vein images are classified based on Image Quality Assessment (IQA) without giving clear indication of the actual IQA used and any experimental results.

Mobile and On-the-Move Acquisition

  • Hand-Based Vascular Traits
  • Eye-Based Vascular Traits

In the context of finger vein recognition, reflected light illumination has been investigated [308] as it is clear that transillumination cannot be implemented in smartphones. In contrast to this simple solution, a custom-built, reflectance-based imaging accessory [239] for obtaining finger veins has been developed.

Disease Impact on Recognition and (Template) Privacy

Motion capture can be ruled out for retinal biometrics as retinal illumination requires a focused and precise illumination process. The extent of privacy-threatening information that can potentially be extracted also depends significantly on the type of data to be analyzed.

Conclusion and Outlook

Even for templates, a representation of the vascular network based on binary structure reveals much more information compared to a representation based on minutiae or even texture properties. For eye-targeted modalities (i.e., retinal and scleral recognition), the future does not appear to be as promising, as many hurdles remain.

Liu F, Yang G, Yin Y, Wang S (2014) Singular value decomposition based minutiae matching method for finger vein recognition. Qin H, Chen Z, He X (2018) Evaluation of finger-vein image quality based on gray-scale and binary image representation. Zhou L, Gongping Yang L, Yang YY, Li Y (2015) Evaluation of finger vein image quality based on support vector regression.

A High-Quality Finger Vein Dataset Collected Using a Custom-Designed

  • Introduction
  • Overview of Finger Vein Acquisition Systems .1 Types of Sensors
    • Commercial Sensors
    • Sensors Developed by Academics
  • University of Twente Finger Vein Capture Device
  • Description of Dataset
  • Results
    • Performance Analysis
  • Next-Generation Finger Vein Scanner .1 Overview
    • Illumination Control
  • Conclusions
  • Future Work

Currently, we are investigating optimal lighting and settings of the cameras and 3D fingerprint reconstruction. In Fig.2.8 a comparison is made between images recorded with the first and second generation finger vein scanner from the University of Twente. Finally, 3D scanning techniques allow compensation for distortions of the finger vein pattern caused by rotation of the finger.

Fig. 2.1 Reflection, transmission and side illumination acquisition
Fig. 2.1 Reflection, transmission and side illumination acquisition

OpenVein—An Open-Source Modular Multipurpose Finger Vein Scanner

Introduction

Both scanners capture high-quality, high-resolution images of finger veins and provide high recognition performance. Use of our proposed scanner design and reproduction of the finger vein scanner itself is free for research purposes. Section 3.3 discusses all the important details and individual parts of our proposed finger vein reader design.

Finger Vein Scanners

  • Light Source Positioning
  • Two Main Perspectives of the Finger—Dorsal and Palmar
  • Commercial Finger Vein Scanners
  • Finger Vein Prototype Scanners and Datasets in Research

The first publicly available finger vein dataset was established in 2008 by Peking University (PKU) [11] using its own prototype scanner (PKU Proto). Our most recent finger vein database is PLUSVein-FV3 [12], captured by the scanner design presented in this chapter (PLUS OpenVein). Figure 3.4 shows some example images for the available finger vein datasets (except for PKU and CFVD).

Fig. 3.1 Light source and image sensor positioning, left: light transmission, right: reflected light.
Fig. 3.1 Light source and image sensor positioning, left: light transmission, right: reflected light.

PLUS OpenVein Finger Vein Scanner

  • Advantages and Differences to Existing Designs
  • Image Sensor, Lens and Additional Filter
  • Light Transmission Illuminator
    • LED-Based Version
    • Laser Module Based Version
  • Reflected Light Illuminator
  • Illuminator Brightness Control Board
    • Arduino Firmware
  • Finger Placement Unit
  • Housing Parts
  • Capturing Software
    • Automated Brightness Control Algorithm

In addition, our scanner housing prevents any direct exposure of the eye to the LED radiation. Due to the narrow beam half angle of the laser modules (note that the laser diodes themselves do not have such a narrow beam angle, instead the focus adjustable lens inside the housing allows for such a small angle), most of the light flux remains within the middle areas of the finger (where most of the veins are located) and so the contrast in these areas remains stable when the finger is moved up (away from the radiator). Therefore, for the current version of the scanner, we recommend the LED-based version to save costs.

Figure 3.5 shows both of the scanners fully assembled and with the right and front side of the scanner half open including labelled parts
Figure 3.5 shows both of the scanners fully assembled and with the right and front side of the scanner half open including labelled parts

PLUSVein-FV3 Finger Vein Dataset

In the meantime, we expanded the dataset to also include palmar finger vein images taken from the same subjects. The second most important reason is that we are interested in participating in expanding our data set and developing it into a comprehensive, open digital finger vein data set available for research purposes. We are currently discussing the possibilities for an appropriate online platform to effectively manage such collaboration and trying to clarify the legal aspects (consent forms must include the right to merge individual data sets together, which of course includes sharing finger vein data with other partners in different countries and in accordance with different legislations) exchanging images of finger veins.

Fig. 3.17 PLUSVein-FV3 example images, top: laser module based scanner, bottom: LED-based scanner
Fig. 3.17 PLUSVein-FV3 example images, top: laser module based scanner, bottom: LED-based scanner

Conclusion

  • Future Work

We are also currently expanding the first version of our already available finger vein dataset internally. Kauba C, Prommegger B, Uhl A (2018) The two sides of the finger - dorsal or palmar - which one is better in finger vein recognition. Prommeger B, Kauba C, Uhl A (2019) Another view on the finger-multi-perspective score level fusion in finger-vein recognition.

Online References and Data Sheets

Microchip Corporation (2018) Microchip AVR ATmega328P 8-Bit Microcontroller Product Page.https://www.microchip.com/wwwproducts/en/ATmega328P. Renesas Electronics (2019) Application Note AN1737 – eye safety for proximity detection using infrared LEDs.https://www.renesas.com/eu/en/doc/application-note/an1737. Texas Instruments Corporation (2018) Texas instruments TLC59401 16-channel LED driver with point correction and grayscale PWM control data sheet.http://www.ti.com/lit/ds/sbvs137/.

An Available Open-Source Vein Recognition Framework

Introduction

The second step is pre-processing to improve the quality of the vein patterns. Another important aspect is that it should be easy to incorporate a new recognition scheme into the existing toolchain to evaluate its recognition performance. For various parts of the recognition toolchain, there are some publicly available implementations available, especially for the feature extraction and comparison step.

Related Work

The rest of this chapter is structured as follows: Section 4.2 discusses related work on publicly available vein recognition software. The vein recognition schemes in our OpenVein toolkit are listed and described in Section 4.4. All those individual components are bundled in an easy-to-use MATLAB-based vein recognition framework, which is available for free.

PLUS OpenVein Toolkit

  • Directory Structure
  • Settings Files
  • External Dependencies

Feature Extraction: most stand-alone feature extraction functions, such as those provided by Bram Ton. The settings are described in more detail in the PLUS OpenVein Toolkit readme file. More details on each individual implementation can be found in the source code files of the respective scheme.

Fig. 4.1 Biometric recognition system
Fig. 4.1 Biometric recognition system

Included Vein Recognition Schemes

  • Input File Handling/Supported Datasets
  • Preprocessing
  • Feature Extraction
  • Comparison
  • Comparison/Evaluation Protocols
  • Performance Evaluation Tools
  • Feature and Score-Level Fusion

Initially, they used Gray Level Grouping to reduce illumination fluctuations and improve contrast in the finger vein images. Then, a circular Gabor filter is used to further improve the visibility of the vein ridges in the images. Since the veins appear as valleys in the cross-sectional profile of the image, RLT [31].

Fig. 4.4 Schematic ISO/IEC 19795-1 compliant ROC (left) and DET plot (right)
Fig. 4.4 Schematic ISO/IEC 19795-1 compliant ROC (left) and DET plot (right)

Experimental Example

  • Dataset and Experimental Set-Up
  • Experimental Results

Comparison: Miura Matcher [32] with horizontal displacement of 80 pixels, vertical displacement of 30 pixels and rotational displacement of 2◦ combined with FVC evaluation protocol [26] is used during the comparison of the extracted vessel features. Figure 4.5 shows an example image of the UTFVP dataset (first column) and the same image after masking the vein region (second column) and after applied preprocessing (third column). Thus, the performance of the two feature extractors and the entire toolchain is not optimized to achieve the best possible recognition performance.

Figure 4.5 shows an example image of the UTFVP dataset (first column) and the same image after vein region masking (second column) and after the applied preprocessing (third column)
Figure 4.5 shows an example image of the UTFVP dataset (first column) and the same image after vein region masking (second column) and after the applied preprocessing (third column)

Conclusion and Future Work

Kauba C, Piciucco E, Maiorana E, Campisi P, Uhl A (2016) Advanced variants of feature level fusion for finger vein recognition. Kauba C, Prommegger B, Uhl A (2018) Focusing the beam – a new laser illumination-based dataset providing insights into finger-vein recognition. Prommegger B, Kauba C, Uhl A (2019) Another view of the finger - multi-perspective score level fusion in finger-vein recognition.

Online Resources

Syarif MA, Ong TS, Teoh AB, Tee C (2017) Improved maximum curvature descriptors for finger vein verification. Xie SJ, Lu Y, Yoon S, Yang J, Park DS (2015) Intensity variation normalization for finger vein recognition using guided filter based singe scale retinex. Zhang J, Yang J (2009) Fingerprint image enhancement based on combination of gray level clustering and circular gabor filter.

Hand and Finger Vein Biometrics

Use Case of Palm Vein Authentication

  • Introduction
  • Palm Vein Sensing
  • Sensor Products with Reflection Method
  • Matching Performance
  • Use Cases of Palm Vein Authentication .1 Usage Situation
    • Login Authentication
    • Physical Access Control Systems
    • Payment Systems
    • Financial Services
    • Health Care
    • Airport Security
    • Government and Municipal
  • Conclusion

Because palm vein authentication has many public uses, it tends to have more recorders than sensors. Moreover, tablets with integrated palm vein authentication have been put into practical use (Fig.5.5) [8]. Palm vein authentication works contactless; this is an optimal feature for public use.

Fig. 5.1 Palm vein image captured by experiment device
Fig. 5.1 Palm vein image captured by experiment device

Evolution of Finger Vein Biometric Devices in Terms of Usability

  • Introduction
    • Early Implementation
    • Commercialisation
    • Evolutions of the Finger Vein Biometric Devices
  • Compliance with Regulations .1 Use Case/Background
    • Usability Requirement Details
    • Challenges
    • Implementation
  • Compactness
    • Use Case/Background
    • Usability Requirement Details
    • Challenges
    • Implementation
  • Portability and Mobility .1 Use Case/Background
    • Usability Requirement Details
    • Challenges
    • Implementation
  • Universal Design .1 Use Case/Background
    • Usability Requirement Details
    • Challenges
    • Implementation
  • Durability and Anti-vandalism .1 Use Case/Background
    • Usability Requirement Details
    • Challenges
    • Implementation
  • High Throughput .1 Use Case/Background
    • Usability Requirement Details
    • Challenges
    • Implementation
  • Universality/Availability .1 Use Case/Background
    • Usability Requirement Details
    • Challenges
    • Implementation
  • Summary

One of the most successful finger vein devices was developed and released by Hitachi, Ltd. To reduce the height of the first generation device H-1, the hood had to be removed. The finger vein device casing has been redesigned to handle outdoor ATM use cases.

Fig. 6.1 Prototype finger vein reader
Fig. 6.1 Prototype finger vein reader

Towards Understanding Acquisition Conditions Influencing Finger Vein

  • Introduction
  • Varying Acquisition Conditions—A Challenging Aspect in Research and Practical Applications
  • Deployed Scanner Devices
  • Finger Vein Acquisition Conditions Dataset
  • Finger Vein Recognition Toolchain and Evaluation Protocol

Their classification is based on the relative positioning of the camera module, the finger and the equipped light source. In the transillumination concept, the light source and the image sensor are placed on opposite sides of the finger taken. The camera is placed on top, the finger can be seen in the middle placed on the finger rest and the lighting module is located at the bottom of the wooden case.

Fig. 7.1 NIR LED and laser-based finger vein scanner (camera on top and finger at bottom)
Fig. 7.1 NIR LED and laser-based finger vein scanner (camera on top and finger at bottom)

Hình ảnh

Table 1.1 Finger vein feature extraction techniques focussing on vascular structure
Table 1.2 Finger vein feature extraction techniques not focussing on vascular structure
Table 1.6 Hand vein datasets available for research (typically upon written request)
Fig. 2.5 Sample images of the left-hand ring finger from the collected dataset
+7

Tài liệu tham khảo

Tài liệu liên quan

Vietnamese state budget revenue should be restructured toward the sustainability and by way of boosting the economy, specifically: (1) Increase the proportion of domestic revenue