Your SlideShare is downloading. ×
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Monocular simultaneous localization and generalized object mapping with undelayed initialization
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Monocular simultaneous localization and generalized object mapping with undelayed initialization

195

Published on

Thesis Defence

Thesis Defence

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
195
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
15
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Monocular Simultaneous Localization and Generalized Object Mapping with Undelayed Initialization 資訊工程所 蕭辰翰1 Robot Perception and Learning Lab 2010/7/23
  • 2. Outline  Introduction  State Vector Definition  Proposed Classification Algorithm  Simulations and Real experiment  Conclusion2 Robot Perception and Learning Lab 2010/7/23
  • 3. Outline  Introduction  State Vector Definition  Proposed Classification Algorithm  Simulations and Real experiment  Conclusion3 Robot Perception and Learning Lab 2010/7/23
  • 4. EKF-based SLAM4 Robot Perception and Learning Lab 2010/7/23
  • 5. Monocular SLAM  Camera as the only sensor  Andrew J. Davison et al. proposed a EKF-based SLAM approach • Andrew J. Davison, Ian Reid, Nicholas Molton and Olivier Stasse: MonoSLAM: Real-Time Single Camera SLAM, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 29, NO. 6, JUNE 2007  Feature states are the 3D position vectors of the locations of point features Multiple images acquired must be combined to achieve accurate depth estimates5 Robot Perception and Learning Lab 2010/7/23
  • 6. Inverse Depth Parametrization6 Robot Perception and Learning Lab 2010/7/23
  • 7. Inverse Depth Parametrization  High degree of linearity  Ability to cope with features far from the camera  Undelayed initialization J.M.M. Montiel, Javier Civera and Andrew J. Davison: “Unified Inverse Depth Parametrization for Monocular SLAM”. Robotics: Science and Systems Conference 2006.7 Robot Perception and Learning Lab 2010/7/23
  • 8. Dynamic Environments  Inclusion of moving features => Degrade performance  Prior knowledge  Avoid moving objects • SomkiatWangsiripitak, David W. Murray: Avoiding moving outliers in visual SLAM by tracking moving objects, ICRA20098 Robot Perception and Learning Lab 2010/7/23
  • 9. Performance Degrade SLAM with static features SLAM with static features and a new moving feature Chieh-ChihWang, Ko-Chih Wang, Chen-Han Hsiao, Kuen-Han Lin and Yi-Liu Chao.: Monocular Vision-based Simultaneous Localization, Mapping and Moving Object Tracking, submit to journal9 Robot Perception and Learning Lab 2010/7/23
  • 10. Classification Stage  Accumulate temporal information  Multiple images required  Abnormal negative inverse depth Chieh-ChihWang, Ko-Chih Wang, Chen-Han Hsiao, Kuen-Han Lin and Yi-Liu Chao.: Monocular Vision-based Simultaneous Localization, Mapping and Moving Object Tracking, submit to journal10 Robot Perception and Learning Lab 2010/7/23
  • 11. Contributions in this thesis  SLAM with generalized objects  Proposed parametrization  Undelayed Initialization  Classification algorithm  based on velocity11 Robot Perception and Learning Lab 2010/7/23
  • 12. Outline  Introduction  State Vector Definition  Proposed Classification Algorithm  Simulations and Real experiment  Conclusion12 Robot Perception and Learning Lab 2010/7/23
  • 13. State vector definition  EKF-based SLAM with generalized objects  State vector:  Camera:  Generalized object:  The inverse depth parametrization  Proposed parametrization  With motion model13 Robot Perception and Learning Lab 2010/7/23
  • 14. Dynamic Inverse Depth Parametrization  9-dimension state value  Position  Velocity  3D Location w.r.t. XYZ coordinate system:14 Robot Perception and Learning Lab 2010/7/23
  • 15. Feature parametrization and motion prediction15 Robot Perception and Learning Lab 2010/7/23
  • 16. Dynamic Inverse Depth Parametrization  Motion Predict  (constant velocity assumption)16 Robot Perception and Learning Lab 2010/7/23
  • 17. Measurement Model  The observation of a point feature17 Robot Perception and Learning Lab 2010/7/23
  • 18. Undelayed Feature Initialization  Initialized using only one image  First observed frame18 Robot Perception and Learning Lab 2010/7/23
  • 19. Initial value of inverse depth and velocity  Initial value of inverse depth  Range of depth: [ d min , ] 1  Range of inverse depth: [ 0 , ] d min  To cover its 95% acceptance region: ˆ 0 1 1 , 2 d min 2 d min  Initial value of velocity  Range of velocity: [ | v | max , | v | max ]  To cover its 95% acceptance region: v 0 | v | max ˆ 0, v 219 Robot Perception and Learning Lab 2010/7/23
  • 20. Outline  Introduction  State Vector Definition  Proposed Classification Algorithm  Simulations and Real experiment  Conclusion20 Robot Perception and Learning Lab 2010/7/23
  • 21. Classification Cue  Conduct simulation to show the convergency of velocity  3 targets in the simulation  Coded in dynamic inverse depth parametrization21 Robot Perception and Learning Lab 2010/7/23
  • 22. Classification Cue: Velocity Convergency22 Robot Perception and Learning Lab 2010/7/23
  • 23. Classification Cue: Velocity Convergency23 Robot Perception and Learning Lab 2010/7/23
  • 24. Classification Cue: Velocity Convergency24 Robot Perception and Learning Lab 2010/7/23
  • 25. Score function for classifying static objects  Given the velocity distribution:  Probability density function value of the velocity distribution at  The relative likelihood at  Classification by thresholding  Cs(X ) ts => classify as static object25 Robot Perception and Learning Lab 2010/7/23
  • 26. Threshold selection on ts  PDF value of static object at is expected higher  Threshold selection t s26 Robot Perception and Learning Lab 2010/7/23
  • 27. Score function for classifying moving objects  Given the velocity distribution:  Mahalanobis distance function  The velocity of moving objects is expected to converge away from  Classification by thresholding  Cm (X ) tm => classify as moving object27 Robot Perception and Learning Lab 2010/7/23
  • 28. Threshold selection on tm  M-dist of a moving object at is expected to larger  Threshold selection t m28 Robot Perception and Learning Lab 2010/7/23
  • 29. Classification State Initialized feature  Generalized objects in state vector  Unknown state Unknown state  Static state  Moving state Cs(X ) ts Cm (X ) tm  Low computational classification Static Moving algorithm state state29 Robot Perception and Learning Lab 2010/7/23
  • 30. State transition Unknown state to Unknown state to Static state Moving state  Change the label  Change the label  Adjust values to  Keep the same values satisfied the property v 0, v 0 SLAM with generalized object is achieved.30 Robot Perception and Learning Lab 2010/7/23
  • 31. Issue on unobservable situations disability of monocular system to find an unique trajectory of an object under the constant-velocity assumption31 Robot Perception and Learning Lab 2010/7/23
  • 32. Issue on unobservable situations  Conduct simulation to show the convergency of velocity  3 targets in the simulation  Coded in dynamic inverse depth parametrization32 Robot Perception and Learning Lab 2010/7/23
  • 33. Ambiguation under unobservable situations33 Robot Perception and Learning Lab 2010/7/23
  • 34. Ambiguation under unobservable situations  Cannot distinguish the state according to the velocity distribution  Ambiguation of  Static object  Constant speed parallel-moving object34 Robot Perception and Learning Lab 2010/7/23
  • 35. Non-parallel moving object under unobservable situations35 Robot Perception and Learning Lab 2010/7/23
  • 36. Non-parallel moving object under unobservable situations  95% confidence region do not cover (0,0,0)  No ambiguation with static object36 Robot Perception and Learning Lab 2010/7/23
  • 37. Classification under unobservable situations  Ambiguation  Static object  Constant speed parallel-moving object  Non ambiguation  Constant speed non parallel-moving object37 Robot Perception and Learning Lab 2010/7/23
  • 38. Outline  Introduction  State Vector Definition  Proposed Classification Algorithm  Simulations and Real experiment  Conclusion38 Robot Perception and Learning Lab 2010/7/23
  • 39. Simulation Observable situation Unobservable situation39 Robot Perception and Learning Lab 2010/7/23
  • 40. Simulation setting  50 Monte Carlo simulations  300 static landmarks and 288 moving landmarks  In each simulation40 Robot Perception and Learning Lab 2010/7/23
  • 41. Classification error ratio Observable situation Unobservable situation41 Robot Perception and Learning Lab 2010/7/23
  • 42. Classification Result  Classification Result of 50 Monte Carlo simulations under threshold  Observable situation42 Robot Perception and Learning Lab 2010/7/23
  • 43. Convergency of SLAM with generalized objects  Convergency of camera43 Robot Perception and Learning Lab 2010/7/23
  • 44. Convergency of SLAM with generalized objects  Convergency of static objects44 Robot Perception and Learning Lab 2010/7/23
  • 45. Convergency of SLAM with generalized objects  Convergency of moving objects45 Robot Perception and Learning Lab 2010/7/23
  • 46. Real Experiment  NTU PAL7 robot  Wide-angle camera  79.48 degree view angle  640 × 480  Laser scanner  At the basement of CSIE  1793 images,13.65fps, 131 seconds46 Robot Perception and Learning Lab 2010/7/23
  • 47. Real Experiment at basement of CSIE47 Robot Perception and Learning Lab 2010/7/23
  • 48. Real Experiment  Classification result48 Robot Perception and Learning Lab 2010/7/23
  • 49. Video of Real Experiment49 Robot Perception and Learning Lab 2010/7/23
  • 50. Comparison of estimation and ground-truth (Topview)50 Robot Perception and Learning Lab 2010/7/23
  • 51. Comparison of estimation and ground-truth (Sideview)51 Robot Perception and Learning Lab 2010/7/23
  • 52. Outline  Introduction  State Vector Definition  Proposed Classification Algorithm  Simulations and Real experiment  Conclusion52 Robot Perception and Learning Lab 2010/7/23
  • 53. Conclusions  Achieve SLAM with generalized objects  Simulations  Real experiments  Adopt un-delayed initialization  Provide a low computational classification algorithm  Competitive performance to laser-based approach53 Robot Perception and Learning Lab 2010/7/23

×