3D掌纹识别：形状指标表示与脆弱位的应用

147 浏览量更新于2024-07-15 收藏 1.53MB PDF 举报

"3D palmprint recognition using shape index representation and fragile bits" 这篇研究论文探讨了3D掌纹识别技术，特别是利用形状指数表示法和脆弱位（fragile bits）来提高识别系统的性能和鲁棒性。在过去的十年里，大多数工作集中在二维（2D）掌纹识别上，但2D图像会丢失形状信息，并且实际应用中易受伪造或噪声干扰，不够稳健。因此，3D掌纹识别被视为一个重要的替代方案，旨在提升现有掌纹识别系统的性能和可靠性。在这项研究中，作者首先通过使用形状指数公式来探索3D掌纹数据的几何信息。形状指数是一种描述曲面形状特征的有效方法，它能够捕捉到3D掌纹的独特轮廓。接着，他们提取了基于高波算子（Gabor wavelet）的特征，高波算子在纹理分析和模式识别领域具有良好的表现，能够捕获掌纹的细节特征。此外，研究者首次发现将脆弱位引入到3D掌纹识别系统中。脆弱位是一种可以反映微小变化的位序列，它们对篡改或噪声非常敏感，从而增强了识别的精度。当掌纹数据受到轻微扰动时，脆弱位的变化可以被用来检测并定位这些变化，这对于防伪和识别准确性至关重要。论文还提到了获取和处理3D掌纹数据的方法，这可能包括使用三维扫描设备来捕捉手掌的立体形状，然后通过图像处理技术提取特征。作者可能还讨论了如何构建和训练识别模型，以及在不同条件下的实验结果，以证明他们的方法在实际应用中的优越性。这篇研究论文为3D掌纹识别提供了一种新的、基于形状指数和脆弱位的方法，该方法有望解决2D掌纹识别的局限性，提高识别的稳定性和抗干扰能力，对于生物识别技术和安全领域的进步具有重要意义。

Suppose the height of the reference plane in Fig. 1 is 0. Through casting structured light

across the grating onto the surface of object, the distance from point C at spatial coordinates (x,

y) to the reference plane could be measured according to [24]:

ox; y

ðÞ

CE ¼

BD  tanα

1 þ tanα

tanβ

ð1Þ

where

BD ¼

2π

λ ð2Þ

Here λ represents the structured light wavelength,α is the casting angle whileβ denotes the

angle between the reference plane and the line passing across the center of CCD and the

current point. φ

denotes the phase difference between points D and E. Since the phase of

point E on the object is equivalent to the phase of point B on the reference plane, φ

is then

computed by:

¼ φ

−φ

ð3Þ

Through projecting several different kinds of structured light patterns (13 patterns are used

in [18]) on the object surface, its depth information can be finally recovered by using the phase

unwrapping and shifting technology [22]andtheformula(1). We recommend the literature

[22] for readers to know more about this technique.

The initial spatial resolution of the acquired 3D palmprint by the developed system [18]is

about 768 × 576 points. However, the redundant parts and noisy boundary areas are useless for

further feature extraction and recognition procedure. The authors in [18] thus utilized a quite

simple region-of-interest (ROI) extraction procedure to obtain the central region of palmprint.

Consequently, the ultimate size of the released 3D palmprint is 128 × 128 points.

3 Feature extraction from 3D palmprint

3.1 Shape index calculation

Once the obtained ROI is available, it is significant to extract effective, reliable and exclusive

features for the further processing. As shown from Fig. 1, the depth information of the captured

3D palmprint in HKPU dataset denotes the relative distance along each point in a certain object

and the reference plane. However, as it was discussed in literature [18, 29], the accuracy of

such relative distance could be deteriorated by random noise and the change of pose when

scanning. Consequently, the variations and fluctuations in the cloud data of 3D palmprint let

the famous iterative closest points (ICP) algorithm [2] unsuitable for 3D palmprint recognition

task. Conversely, surface curvature maps paradise, such as mean or Gaussian curvatures of the

principal lines and strong wrinkles, which could exclusively portray the geometric structure of

a surface, are known quite efficient and reliable in depicting the unique characteristics of a

certain 3D palmprints [30].

Assume that, on the surface patch Z,P is a point to be evaluated. L

cur

denotes all passed

curves on Zthat cross the evaluated point P. With regard to each curve, it will has an relevant

curvature L

. For these curvatures L

, there exist at least one maximal curvature l

and one

15360 Multimed T ools Appl (2017) 76:15357–15375

剩余19页未读，继续阅读

weixin_38632046

粉丝: 10
资源: 933

3D掌纹识别：形状指标表示与脆弱位的应用

掌纹识别方法综述A Comparative Study of Palmprint Recognition Algorithms）

Palmprint Recognition Using PhaseCongruency and Region-Based Orientation Code

Joint deep convolutional feature representation for hyperspectral palmprint recognition

Local apparent and latent direction extraction for palmprint recognition

Fusion of Contour Feature and Edge Texture Information for Palmprint Recognition

Combine crossing matching scores with conventional matching scores for bimodal biometrics and face and palmprint recognition experiments

Histogram of Oriented Lines for Palmprint Recognition

Precision direction and compact surface type representation for 3D palmprint identification

Palmprint Recognition System Matlab Code:掌纹识别系统-开源

Palmprint identification using restricted fusion

最新资源