Towards compressing SLAM maps for 6D relocalisation

We develop and evaluate a novel data compression strategy for visual SLAM centered on traveled trajectory analysis. Beyond compressing scene structure based purely on geometrical information as commonly done, we aim at developing representations that are useful for re-exploration, that use low memory footprint and low CPU computation. Our work is evaluated on data collected from a visual sensor and exploits the information intrinsic to the trajectory of exploration together with the visual information of map points. We perform rigorous statistical evaluation and Pareto analysis to show how this approach compares with a traditional keyframe-based data representation and the compromises achievable in terms of compression rate and relocalisation success.