Single-photon dehazing imaging method based on density clustering-guided Gaussian model fitting.

Area of Science:

• Photonics and Optical Sensing
• Computational Imaging
• Environmental Monitoring

Background:

• Single-photon counting lidar struggles in smoke due to scattering noise and low signal-to-noise ratios (SNRs).
• Residual noise photons after Gamma fitting hinder accurate target Gaussian model fitting in smoke.
• Existing methods have limited performance in smoke-penetrating 3D imaging.

Purpose of the Study:

• To develop a robust smoke-penetrating 3D imaging algorithm for challenging scattering environments.
• To improve the accuracy and quality of lidar imaging in smoke-filled conditions.
• To effectively separate target signals from residual noise photons.

Main Methods:

• A Gamma model was used for initial echo photon data fitting and backscattering peak separation.
• DBSCAN (Density-Based Spatial Clustering of Applications with Noise) density clustering analyzed fitting residuals to remove noise photons.
• Gaussian model fitting was applied to filtered photons for precise depth estimation.

Main Results:

• The proposed algorithm demonstrated superior performance on the Middlebury simulation dataset, showing lower RMSE and higher SSIM across various smoke particle sizes.
• Real-world experiments in rainy and foggy conditions validated the algorithm's reconstruction capabilities.
• Significant enhancement in imaging quality and accuracy in smoke-scattering environments was achieved.

Conclusions:

• The DBSCAN residual clustering-guided Gaussian model fitting algorithm effectively enhances smoke penetration for lidar imaging.
• This method offers a significant advancement for 3D imaging in adverse atmospheric conditions.
• The algorithm provides a reliable solution for accurate depth estimation and object reconstruction in scattering environments.