1 / 19

SLAM – Loop Closing with Visually Salient Features

SLAM – Loop Closing with Visually Salient Features. Paul Newman, Kin Leong Ho Oxford University Robotics Research Group. Motivation. Loop Closing – the task of deciding whether a vehicle has returned to a previously visited area

stacie
Download Presentation

SLAM – Loop Closing with Visually Salient Features

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. SLAM – Loop Closing with Visually Salient Features Paul Newman, Kin Leong Ho Oxford University Robotics Research Group

  2. Motivation • Loop Closing – the task of deciding whether a vehicle has returned to a previously visited area • Popular approaches – nearest neighbour statistical gate, joint compatibility test • Closing loop with visually salient features to avoid dependence on global position estimate

  3. Visual Saliency • Scale saliency detector [ Kadir/Brady IJCV 2001] • - form p.d.f. of pixel properties within local region at varying scales for each pixel • detection of region at a particular scale where weighted entropy is peaked • selected regions are considered more “interesting” p.d.f of local pixels overscale s around position x entropy

  4. Visual Saliency Dissimilarity p.d.f of across scale Weighting Function Saliency metric entropy Weighting function

  5. Wide-Baseline Stability • Maximally stable extremal region (MSER) detector [ Matas etal. BMVC 2002] • -pixels taking on values in the range D = {dmin ….dmax}

  6. MSER detector Saliency detector

  7. Feature Description -Scale invariant feature transform (SIFT) descriptor [David Lowe IJCV 2004] -128 dimensional descriptor

  8. MSER Detector SIFT Descriptor Query Image Selected Regions Saliency Detector Similarity Measure Matched Images Laser Scan Database Image Database

  9. Demonstration of wide-baseline stability of visually salient features under perspective distortion and variation in illumination conditions

  10. Matching Performance Similar posters found in the environment.

  11. A Delayed State Formulation Control Past poses Scan matching between Past poses produces observation z with which to update state-vector State vector contains only past vehicle poses. (Atlas IJRR 2004 )

  12. Delayed State Formulation II EKF update

  13. Closing Small Loops

  14. Closing Big Loops

  15. Closing the loop

  16. Issues -hard decision making -using saliency detector as binary selector -repetitive visual features in urban environment

  17. Demonstration

  18. Questions Thank you!

More Related