Joint surface detection and depth estimation from single-photon Lidar data using ensemble estimators

Abstract

This paper addresses the problem of joint surface detection and depth estimation from single-photon Lidar (SPL) data. Traditional 3D ranging methods for SPL usually perform surface detection and range estimation sequentially to alleviate the computational burden of joint detection and estimation. Adopting a Bayesian formalism, the joint detection/estimation problem is formulated as a single inference problem. To avoid the intractable integrals usually involved with variable marginalization, we consider discrete variables and the resulting problem is recast as a model selection/averaging problem.We illustrate our method for a case where the expected signal-to-background (e.g., the target reflectivity and ambient illumination level) is unknown but the proposed framework can be adapted to more complex problems where the target depth can be obtained by combining several estimators. We demonstrate the additional benefits of the proposed method in also providing a conservative approach to uncertainty quantification of the calculated depth estimates, which can be used for real time analysis. The benefits of the proposed methods are illustrated using synthetic and real SPL data for targets at up to 8.6 km.