• Acta Photonica Sinica
  • Vol. 53, Issue 9, 0910004 (2024)
Yilong ZHANG, Shengming ZHU, Haixia WANG*, Haohao SUN, and Rui YAN
Author Affiliations
  • School of Computer and Science Technology, Zhejiang University of Technology, Hangzhou 310000, China
  • show less
    DOI: 10.3788/gzxb20245309.0910004 Cite this Article
    Yilong ZHANG, Shengming ZHU, Haixia WANG, Haohao SUN, Rui YAN. OCT Internal and External Fingerprint Extraction Method Based on Soft Label[J]. Acta Photonica Sinica, 2024, 53(9): 0910004 Copy Citation Text show less
    Optical schematic of SD-OCT system[27]
    Fig. 1. Optical schematic of SD-OCT system27
    OCT fingerprint data
    Fig. 2. OCT fingerprint data
    Overall process of the algorithm
    Fig. 3. Overall process of the algorithm
    Labeling process
    Fig. 4. Labeling process
    Coordinate attention block[14]
    Fig. 5. Coordinate attention block14
    Network architecture of SLCA-UNet
    Fig. 6. Network architecture of SLCA-UNet
    External fingerprint images generated by different algorithms
    Fig. 7. External fingerprint images generated by different algorithms
    Internal fingerprint images generated by different algorithms
    Fig. 8. Internal fingerprint images generated by different algorithms
    The DET curves of external and internal fingerprint
    Fig. 9. The DET curves of external and internal fingerprint
    Fingerprint reconstruction step based on the soft Label
    Input. B-scan: I(x,y,z),the predict result of the stratum corneum layer and viable epidermal layer: Pexternalx,y,z),Pinternalx,y
    Output. External fingerprint and Internal fingerprint
    Step1. x=1 800,y=500,z=1 400,t=1
    Step2. Initialize:External fingerprint =zeros([zx])Internal fingerprint=zeros([zx])
    Step3. for t in range(1,z+1):(1-1400 B-scan image)
    Step4. finger_external[t-1,:]=sum(I(:,:,t)*Pexternal(:,:,t),axis=0)/(sum(Pexternal(:,:,t),axis=0))
    Step5. finger_internal[t-1,:]=sum(I(:,:,t)*Pinterna(:,:,t),axis=0)/(sum(Pinternal(:,:,t),axis=0))
    Step6. t=t+1 (process the next B-scan image)
    Step7. External fingerprintl=normalize(External fingerprint), the external fingerprint was normalized to 0 to 255
    Step8. Internal fingerprint=normalize(Internal fingerprint), the internal fingerprint was normalized to 0 to 255
    Table 1. Fingerprint reconstruction steps based on soft Label
    Extraction methodFingerprintMean values of NFIQ 2.0 score
    ZAM 1External fingerprint18
    Internal fingerprint36
    ZAM 2External fingerprint28
    Internal fingerprint35
    ZAM 3External fingerprint14
    Internal fingerprint35
    U-NetExternal fingerprint31
    Internal fingerprint41
    GradientExternal fingerprint32
    Internal fingerprint42
    SLCA-UNetExternal fingerprint35
    Internal fingerprint47
    Table 2. NFIQ scores of different algorithms
    Extraction methodDenoising time/sExtraction time/s
    ZAM176 00091
    U-Net280127
    Gradient16 000580
    SLCA-UNet0140
    Table 3. Time comparison of different algorithms
    GroupMethodNFIQ 2.0 scoreEER scoreExtraction time/s
    External fingerprintInternal fingerprintExternal fingerprintInternal fingerprint
    1U-Net31411.1%1.5%127
    2UNet+Coordinate attention31441.2%1.2%140
    3U-Net+Soft Label34460.9%0.9%127
    4Our method(UNet+Soft Label+Coordinate attention)35470.8%0.8%140
    Table 4. Results of the ablation experiments of each module
    GroupMethodNFIQ 2.0 scoreEER score
    External fingerprintInternal fingerprintExternal fingerprintInternal fingerprint
    1Smooth contour boundary Label (our method)35470.8%0.8%
    2Unsmooth contour boundary Label253522.5%0.9%
    Table 5. Experimental results for the contour boundaries are smooth or not
    Yilong ZHANG, Shengming ZHU, Haixia WANG, Haohao SUN, Rui YAN. OCT Internal and External Fingerprint Extraction Method Based on Soft Label[J]. Acta Photonica Sinica, 2024, 53(9): 0910004
    Download Citation