Dense monocular Simultaneous Localization and Mapping by direct surfel optimization

Abstract

This work presents a novel approach for monocular dense Simultaneous Localization and Mapping. The surface to be estimated is represented as a piecewise planar surface, defined as a group of surfels each having as parameters the position and normal. These parameters are directly estimated from the raw camera pixels measurements using a Gauss-Newton iterative process. The representation of the surface as a group of surfels has many advantages. First, it allows recovering robust and accurate pixel depths, without the need to use a computationally demanding depth regularization schema. This has the further advantage of avoiding the use of a physically unlikely surface smoothness prior. What is more, new surfels can be correctly initialized from the information present in nearby surfels, avoiding also the need to use an expensive initialization routine commonly needed in Gauss-Newton methods. The method was written in the GLSL shading language, allowing the use of GPU devices and achieve real-time processing. The method was tested on benchmark datasets, showing both its depth and normal estimation capacity, and its quality to recover the original scene. Results presented in this work showcase the usefulness of the more physically grounded piecewise planar scene depth prior, instead of the more commonly pixel depth independence and smoothness prior.