Crosstalk-induced artifacts and resolution degradation in 3D Geiger-mode lidar data

Geiger-mode (GmAPD) Lidar sensors use 2D avalanche photodiode arrays that record single-photon arrivals with nanosecond-scale precision or better. These GmAPD sensors can operate in otherwise untenable low-power regimes, utilizing coincidence processing algorithms to recover the 3D terrain from the noisy measurements. However, these sensors are also susceptible to crosstalk, a phenomenon whereby avalanching detector pixels generate photons that scatter inside the detector and trigger neighboring pixels to also avalanche. Unlike dark current, these phantom detections are correlated with the signal and cannot be removed by averaging. This phenomenon reduces the spatial resolution of the derived imagery both along range and perpendicular to the line of sight. We utilize synthetic GmAPD lidar point clouds to characterize the effect of crosstalk events on 3D imagery. We qualitatively and quantitatively evaluate these point clouds to compare the changes in standard resolution measurements due to the increasing presence of crosstalk points. We also investigate what terrain features are most corrupted by crosstalk noise in these point clouds. We discuss how these various findings could allow crosstalk noise to be mitigated from a data-based perspective.