Textured occupancy grids for monocular localization without features

Abstract

Abstract-A textured occupancy grid map is an extremely versatile data structure. It can be used to render humanreadable views and for laser rangefinder localization algorithms. For camera-based localization, landmark or featurebased maps tend to be favored in current research. This may be because of a tacit assumption that working with a textured occupancy grid with a camera would be impractical. We demonstrate that a textured occupancy grid can be combined with an extremely simple monocular localization algorithm to produce a viable localization solution. Our approach is simple, efficient, and produces localization results comparable to laser localization results. A consequence of this result is that a single map representation, the textured occupancy grid, can now be used for humans, robots with laser rangefinders, and robots with just a single camera.

6D Relocalisation for RGBD Cameras Using Synthetic View Regression

“…Some representations for these models include overlapping keyframes [15], surfel clouds [9], volumetric signed distance functions [14], and textured occupancy grids [12]. These representations are all capable of generating 3D RGB point clouds, or synthesising novel RGBD views over a wide range of viewpoints by projecting the underlying model into a virtual camera.…”

“…Similarly, view-dependent geometric descriptors, such as VFH [17], can be generated from real and synthetic depth images and used in the same way. Yet another approach is particle filter estimation of pose using online synthesis of views from the map [12]. Typically, the comparison of views is relatively fast in these methods, but online view synthesis can be slow, particularly if graphics hardware is unavailable or if the underlying map representation is complex.…”

“…Then, in the following sections, we present a new view-based method that estimates camera pose from a set of synthetic views. In contrast to related methods, such as [7,11,12], this method uses synthetic RGBD images to estimate full 6D pose at framerate using a regression framework.…”

Monte Carlo Localization using 3D texture maps

“…Recent work has demonstrated that a texture map can be generated by either fusing a camera and a 3D laser range finder [9] or using a RGBD sensor, such as Kinect TM (Microsoft Co., Redmond, WA, USA). However, few methods [10] have discussed the technique of localizing in a texture map.…”

OctoMap: an efficient probabilistic 3D mapping framework based on octrees

“…As an example, we extended our mapping framework to store the average color of each voxel. This creates visualizations for the user and enables a color-based classification of the environment or appearance-based robot localization from virtual views (similar to (Einhorn et al, 2011;Mason et al, 2011)). It can also be used as a starting point to create colored, high-resolution surface meshes (Hoppe et al, 1992).…”