Presentation is loading. Please wait.

Presentation is loading. Please wait.

«CO-REGISTRATION OF HETEREGENEOUS GEOREFERENCING 3D DATA : CONTRIBUTION OF MOBILE POINT CLOUDS CORRECTION » Robotics Lab,

Similar presentations


Presentation on theme: "«CO-REGISTRATION OF HETEREGENEOUS GEOREFERENCING 3D DATA : CONTRIBUTION OF MOBILE POINT CLOUDS CORRECTION » Robotics Lab,"— Presentation transcript:

1 «CO-REGISTRATION OF HETEREGENEOUS GEOREFERENCING 3D DATA : CONTRIBUTION OF MOBILE POINT CLOUDS CORRECTION » Robotics Lab, Mines ParisTech July 2010 Dr. Taha Ridene

2 I.Introduction II.Rigid registration algorithms III.Results of registration IV.Conclusion Outlines 2/19

3 I. 3D data production II. Exploitation of 3D Video gameTouristic Military Acquisition Data processing Digitalizing of the territories and their resources and exploitation of multimedia information 17 partners : 7 companies and 10 public research labs GPS 3/19 Global context 3D Data sets Need for correction Introduction Rigid registration algorithmsResults of registrationConclusion

4 To obtain a 3D mapping database : textured, triangulated and geo-referenced Approach: 3D heteregenous representation fusion 4/19 I. 3D data production Acquisition Data processing Global context 3D Data sets Need for correction Introduction Rigid registration algorithmsResults of registrationConclusion Digitalizing of the territories and their resources and exploitation of multimedia information 17 partners : 7 companies and 10 public research labs

5 5/19 Global context 3D Data sets Need for correction Introduction Rigid registration algorithmsResults of registrationConclusion Mensi Trimble Mines ParisTech IGN

6 3D heteregenous data geo referenced in Lambert2 (1 St initialization of the registration) Coherent datasets Registration/correction Data fusion layout Fusion 6/19 Global context 3D Data sets Need for correction Introduction Rigid registration algorithmsResults of registrationConclusion

7 3D point clouds DSM P1 P2 P3 T2 T3 Portion of Interest T1 P1ref Reference portion Input /19 Correction using DSM ICP-SA R-ICP Introduction Rigid registration algorithms Results of registrationConclusion HPS-ICP

8 Top Convergence criteria MQ M’ Q’ Pre-processing (s) Itératif step 8/19 Correction using DSM ICP-SA R-ICP Introduction Rigid registration algorithms Results of registrationConclusion HPS-ICP KD-Tree acceleration (Mount and Sunil., 2006) Point to point Point to surface Least mean square (LMS) (William et al., 1988) Dynamic threshold (Ridene and Goulette, 2008, CTL 2009, RFPT 2010)

9 Classical ICP-SA meet difficulty Help the algorithm from the start and fails 9/19 Correction using DSM ICP-SA R-ICP Introduction Rigid registration algorithms Results of registrationConclusion HPS-ICP

10 M Q K.N.N computing ICP-SA Initialisation by RANSAC 10/19 Correction using DSM ICP-SA R-ICP Introduction Rigid registration algorithms Results of registrationConclusion HPS-ICP RANSAC (RANdom Sample Concensus) (Fischler and Bolles,1981) In registration (Chen et al., 1999; Bae and Lichti, 2008) (Ridene and Goulette, CIRA 2009)

11 MQ M’ Q’ Horizontal plan extraction Tz estimation Apply T-init 2D projection T2D estimation ICP-SA Initialization by horizontal plan segmentation and registration 11/19 Correction using DSM ICP-SA R-ICP Introduction Rigid registration algorithms Results of registrationConclusion HPS-ICP (Jebbari et al, 2009)

12 MQ M’M’ Q’ Horizontal plan extraction Tz estimation Apply T-init 2D projection T2D estimation ICP-SA Initialization by horizontal plan segmentation and registration RANSAC Segmentation Algorithm « Profile-Based » segmentation 3D 2D DoG Filter 12/19 Correction using DSM ICP-SA R-ICP Introduction Rigid registration algorithms Results of registrationConclusion HPS-ICP

13 MQ M’M’ Q’ Horizontal plan extraction Tz estimation Apply T-init 2D projection T2D estimation ICP-SA Initialization by horizontal plan segmentation and registration 13/19 Correction using DSM ICP-SA R-ICP Introduction Rigid registration algorithms Results of registrationConclusion HPS-ICP

14 14/19 Global results Global performance IntroductionRigid registration algorithms Results of registration Conclusion

15 15/19 IterationCPU Time (s)Acceleration factor Without KD- Tree KD-tree P2/DSM 33620,818,0834,2 P8/DSM 3518,087,4931,3 P1/P2ref ,5234,832,74 Acceleration factor ~32 Intel(R) Xeon (R) CPU 2.00GHZ avec 2Go de RAM Traited area: ~ 5 millions of point Global time = 3 mn Global results Global performance IntroductionRigid registration algorithms Results of registration Conclusion

16 Rigid Registration : Possible solution for Mobile Mapping Systems geo-referencing/localization problems Shift problems Correspondence after registration 16/19 IntroductionRigid registration algorithmsResults of registrationConclusion

17 Publications 17/19 Journals 1.T. Ridene and F. Goulette. Coregistration of DSM and 3D point clouds acquired by a mobile mapping system. Geodetic sciences bulletin - Special Issue on Mobile Mapping Technology, 15(5) : , 2009d. 2.T. Ridene and F. Goulette. Recalage de relevés laser fixes et mobiles sur MNS pour la cartographie numérique 3D. Revue Française de photogrammétrie et de télédétection, Jan. 2009a. International conferences 1.T. Ridene and A. Manzanera. Mécanismes d’attention visuelle sur rétine artificielle. TAIMA’07., Hammamet, May T. Ridene and F. Goulette. Recalage hétérogène de données 3D d’environnements urbains. MajecSTIC’08 (IEEE France), Oct T. Ridene and F. Goulette. Recalage de relevés laser fixes et mobiles sur MNS pour cartographie numérique 3D. Colloque Techniques Laser Pour l’Etude des Environnements Naturels et Urbains, Jan T. Ridene and F. Nashashibi. Localisation précise d’un système de cartographie mobile pour la numérisation 3D d’environnement Urbain. ATEC-ITS, Feb T. Ridene and F. Goulette. Registration together and to DSM of several 3D point clouds issued from a Mobile Mapping System. Mobile Mapping Technologies, jul. 2009b. 6.T. Ridene and F. Goulette. Registration of fixed-and-mobile- based terrestrial laser data sets with DSM. pages , dec. 2009c. doi : /CIRA T. Ridene and F. Goulette. Feature-based quality evaluation of 3D heterogeneous data registration. In Proceedings of SPIE, volume 7526, page 75260Z, T. Ridene and F. Goulette. USAGE DE LA CARTOGRAPHIE 3D POUR L’URBANISME ET LE SERVICE DE PROXIMITÉ -EXEMPLE D’APPLICATION AU DIAGNOSTIC D’ACCESSIBILITÉ. (Accepted) GEOTUNIS 2010 IJCV special ISSUE Sptember 2010

18 T H A N K Y O U F O R Y O U R A T T E N T I O N 18/19

19 14/19 Global results Global performance IntroductionRigid registration algorithms Results of registration Conclusion


Download ppt "«CO-REGISTRATION OF HETEREGENEOUS GEOREFERENCING 3D DATA : CONTRIBUTION OF MOBILE POINT CLOUDS CORRECTION » Robotics Lab,"

Similar presentations


Ads by Google