3D Dense Mapping with the Graph of Keyframe-Based and View-Dependent Local Maps

Abstract This article concerns the problem of a dense mapping system for robot exploring new environment. In this scenario, equipped with an RGB-D camera uses RGB and range data to build consistent model Firstly, requires selection representation. Secondly, has deal localization drift which can be corrected when loop closure is detected. article, we both these problems, make several technical contributions. We define local maps use Normal Distribution Transform (NDT) stored in 2D structures represent scene varying 3D resolution. method directly utilizes uncertainty sensor provides information about accuracy map. also propose architecture that pose covisibility graphs correct global environment after detection. show how integrate graph keyframes management ORB-SLAM2 localization. Finally, advantages view-dependent over methods uniformly divide space objects