Infrared and Laser Engineering, Volume. 54, Issue 6, 20240587(2025)

Photon-efficient lidar signal processing methods based on adaptive denoising module

Zixun WANG1,2,3,4 and Bo LIU1,2,3,4
Author Affiliations
  • 1Institute of Optics and Electronics, Chinese Academy of Sciences, Chengdu 610209, China
  • 2Key Laboratory of Science and Technology on Space Optoelectronic Precision Measurement, Chinese Academy of Sciences, Chengdu 610209, China
  • 3National Key Laboratory of Optical Field Manipulation Science and Technology, Chinese Academy of Sciences, Chengdu 610209, China
  • 4University of Chinese Academy of Sciences, Beijing 100049, China
  • show less
    Figures & Tables(11)
    Comparison of data with same SNR but different accumulation times
    Overall network structure
    (a) Network structure of denoising block;(b) Network structure of self-attention mechanism
    Comparison of histogram processing results for photon histograms at different SNR. (a1)-(a2) Signal-to-noise ratio of 2∶10; (b1)-(b2) Signal-to-noise ratio of 2∶20; (c1)-(c2) Signal-to-noise ratio of 2∶50
    (a) Single photon lidar system equipment diagram; The photon histogram of the input from real scene (b1) and the output echo wave after network processing (b2); (c) Real scene to get data
    The effect of the number of denoising modules on training RMSE
    • Table 1. Comparison of quantitative results in simulated dataset with SNR of 2∶10

      View table
      View in Article

      Table 1. Comparison of quantitative results in simulated dataset with SNR of 2∶10

      MethodsRMSE/mAccuracy(δ=1.01)Accuracy(δ=1.02)Accuracy(δ = 1.03)
      Maximum4.65270.39610.43480.4505
      MLE0.79410.86740.92430.9442
      Correlation0.79430.86790.92400.9445
      Proposed0.20490.91410.95250.9662
      P_AS0.01280.92540.96250.9750
    • Table 2. Comparison of quantitative results in simulated dataset with SNR of 2∶20

      View table
      View in Article

      Table 2. Comparison of quantitative results in simulated dataset with SNR of 2∶20

      MethodsRMSE/mAccuracy(δ=1.01)Accuracy(δ=1.02)Accuracy(δ=1.03)
      Maximum5.19360.27880.30530.3173
      MLE2.16480.70150.75500.7737
      Correlation2.16500.69960.75380.7732
      Proposed0.28270.89960.94210.9564
      P_AS0.01600.91400.95600.9693
    • Table 3. Comparison of quantitative results in simulated dataset with SNR of 2∶50

      View table
      View in Article

      Table 3. Comparison of quantitative results in simulated dataset with SNR of 2∶50

      MethodsRMSE/mAccuracy(δ=1.01)Accuracy(δ=1.02)Accuracy(δ=1.03)
      Maximum4.33450.19380.21390.2206
      MLE3.11750.43060.46310.4688
      Correlation3.11720.43190.46310.4688
      Proposed2.84000.42130.45880.4710
      P_AS0.29100.71920.78330.8042
    • Table 4. Comparison of quantitative results in real dataset using different methods

      View table
      View in Article

      Table 4. Comparison of quantitative results in real dataset using different methods

      MethodsRMSE/m
      Maximum64.8715
      MLE5.7056
      Correlation5.7054
      Proposed/P_AS1.3659/0.6310
    • Table 5. Comparison of ablation experiment results (SNR=2∶10)

      View table
      View in Article

      Table 5. Comparison of ablation experiment results (SNR=2∶10)

      Experimental variablesRMSE/m
      Wo non-local0.3066
      Wo CE0.2215
      Wo MSE0.7656
      Wo OR0.3545
      Proposed0.2049
    Tools

    Get Citation

    Copy Citation Text

    Zixun WANG, Bo LIU. Photon-efficient lidar signal processing methods based on adaptive denoising module[J]. Infrared and Laser Engineering, 2025, 54(6): 20240587

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Laser

    Received: Jan. 24, 2025

    Accepted: Jan. 23, 2025

    Published Online: Jul. 1, 2025

    The Author Email:

    DOI:10.3788/IRLA20240587

    Topics