High-Precision Depth Image Estimation for Array Gm-APD LiDAR Based on Dual-Parameter Model Feature in Dynamic Atmospheric Obscurant Environments.

The nonstationary distribution of dynamic atmospheric obscurants and intense backscattering interference jointly create a severely photon-starved regime, substantially degrading the depth imaging performance of array Gm-APD LiDAR in highly scattering environments. Here, we present a depth imaging estimation algorithm through dynamic atmospheric obscurants, which enables the discrimination of atmospheric obscurant interference and significantly improves depth imaging accuracy. The proposed method employs a three-step strategy comprising data preprocessing, adaptive identification of interference-source regions, and continuous multi-frame depth image fusion based on temporal correlation, thereby enabling efficient suppression of dynamic noise and improved target integrity. The proposed method is successfully demonstrated under different attenuation lengths and dynamic atmospheric obscurant conditions. Across all tested conditions, the proposed method achieves a target recovery rate (TR) ranging from 0.71 to 0.89, a root mean square error (RMSE) ranging from 35.62 to 49.20 time bins (equivalent to 5.34-7.38 m), and a structural similarity (SSIM) ranging from 0.89 to 0.94. Compared with traditional methods, the proposed method improves TR by at least 0.41 (116.2%) and SSIM by at least 0.06 (6.8%), while reducing RMSE by at least 11.03 time bins (23.6%). In particular, under the most challenging condition, with an average attenuation length of 2.43 and an occlusion ratio of 48%, the proposed algorithm achieves a TR of 0.89, an RMSE of 49.20 time bins (equivalent to 7.38 m), and an SSIM of 0.89. These results demonstrate the considerable potential of the proposed method for depth imaging in extremely strong scattering environments.