Fast visual odometry and mapping from RGB-D data

Abstract

An RGB-D camera is a sensor which outputs color and depth and information about the scene it observes. In this paper, we present a real-time visual odometry and mapping system for RGB-D cameras. The system runs at frequencies of 30Hz and higher in a single thread on a desktop CPU with no GPU acceleration required. We recover the unconstrained 6-DoF trajectory of a moving camera by aligning sparse features observed in the current RGB-D image against a model of previous features. The model is persistent and dynamically updated from new observations using a Kalman Filter. We formulate a novel uncertainty measure for sparse RGD-B features based on a Gaussian mixture model for the filtering stage. Our registration algorithm is capable of closing small-scale loops in indoor environments online without any additional SLAM back-end techniques.

Efficient generation of 3D surfel maps using RGB–D sensors

“…A characteristic feature of the approach proposed by Dryanovski et al (2013) is that the ICP algorithm does not work on two most recent points clouds, but instead aligns the newest point cloud to the complete model of the scene constructed from point clouds collected so far. In order to do it efficiently, only selected keypoints from each frame are taken into account and incorporated into the model (the full clouds can be stored in external memory).…”

“…An alternative solution for maintaining map consistency is to match the current frame features with the full map constructed so far, instead of matching them with features of one or several most recent frames (Dryanovski et al, 2013). Efficiency is ensured by storing in the map only selected keypoints from each frame.…”

“…The second novelty concerns the utilization of a visual odometry method based on a sparse ICP algorithm (Dryanovski et al, 2013) that is more efficient than the one proposed in another…”

“…• fast visual odometry and mapping from RGB-D data (Dryanovski et al, 2013) (The City University of New York).…”

“…replacement of a dense ICP algorithm by a much more efficient sparse-ICP method, as proposed by Dryanovski et al (2013).…”