1 / 2

Realtime Visual and Point Cloud SLAM Nicola Fioraio, Kurt Konolige

Realtime Visual and Point Cloud SLAM Nicola Fioraio, Kurt Konolige. Real-Time Visual and Point Cloud SLAM. Goal: Integrate ICP and visual features in frame-frame matching Perform global optimization over all frames and features Do it in real time Techniques for real-time ICP:

papina
Download Presentation

Realtime Visual and Point Cloud SLAM Nicola Fioraio, Kurt Konolige

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Realtime Visual and Point Cloud SLAM Nicola Fioraio, Kurt Konolige

  2. Real-Time Visual and Point Cloud SLAM • Goal: • Integrate ICP and visual features in frame-frame matching • Perform global optimization over all frames and features • Do it in real time • Techniques for real-time ICP: • Subsample on a regular grid (~1000 points) • Fast matching via reprojection • NLSQ constraints, using the cost function ; emulates Segal et al. “Generalized-ICP” (RSS 2009) • is the distance between matching points • defines the non-isotropic error for point-point, point-plane and plane-plane match • Techniques for Bundle Adjustment: • Both ICP and visual features are NLSQ constraints, so they can be used together • Global optimization is Bundle Adjustment over all ICP matches and visual features • Henry et al. “Rgb-d mapping” (ISER 2010) use reduction to pose graph, no features • Endres et al. “RGBD-SLAM” also reduces to pose-pose constraints Realtime performance for the pairwise alignment Global BA: 77899 edges => 346ms • Techniques for Bundle Adjustment: • Both ICP and visual features are NLSQ constraints, so they can be used together • Global optimization is Bundle Adjustment over all ICP matches and visual features • Henry et al. “Rgb-d mapping” (ISER 2010) use reduction to pose graph, no features • Endres et al. “RGBD-SLAM” also reduces to pose-pose constraints

More Related