Computer Vision News - October 2019

7 GSLAM The paper also compared the suggested method against two state of the art SLAM methods. In the following figure you can see the reconstructed trajectory: the suggested method (red dots) ORB-SLAM (blue dots) and LSD- SLAM (orange dots); the yellow and green dots are the ground truth starting and ending points of the trajectory. Results The first set of results from the paper is a visualization of the algorithm. The following figure demonstrates the pose graph generated from key- frames. The edges of different colors indicate three kinds of keyframes; blue for direct neighbors, orange for extended neighbors and red for loops. This pose graph enables a global optimization step (as explained) that improves the final solution. Additional results of this kind can be seen in the following figure: “GSLAM introduces a novel global approach for solving Simultaneous Localization and Mapping problems as opposed to most of the current meth- ods which use incremental schemes”.