融合先验信息到三维重建组会报告[2]

Size: px

Start display at page:

Download "融合先验信息到三维重建组会报告[2]"

유진 류(유)
5 years ago
Views:

1 [1] Crandall D, Owens A, Snavely N, et al. "Discrete-continuous optimization for large-scale structure from motion." (CVPR), 2011 [2] Crandall D, Owens A, Snavely N, et al. SfM with MRFs: Discrete-Continuous Optimization for Large-Scale Structure from Motion.(PAMI),

Andrew Owens CSAIL MIT Department of Computer Science Noah Snavely and

2 Discrete-Continuous Optimization for Large-scale Structure from Motion David Crandall School of Informatics and Computing Indiana University Andrew Owens CSAIL MIT Department of Computer Science Noah Snavely and Dan Huttenlocher Cornell University Runner-up best paper at CVPR 2011

5 MRFBP

6 MRFBP

7 Structure from Motion p 1 p 4 p 2 p 3 minimize f(r,t,x) p 5 p 6 p 7 Camera 1 R 1,t 1 R 2,t 2 Camera 2 Camera 3 R 3,t 3

8 Reconstruction pipeline ANN 3D (IBA) Start with seed model Run bundle adjustment Remove outliers Add another image Repeat R, T P, X

9 MRFBP

10 1SBA scalability 2 IBA[Incremental Bundler Adjustment] 3

11 SFMMRF SFM

12 SFMMRF : LBPloopy belief propagation MRF bundle adjustment

13 SFMMRF IBA 6

14 MRFBP

15 Compute relative pose between camera pairs using 2-frame SfM t ij R ij

16 R ij t ij (Ri, ti)(rj, tj) :

17 GPS tilt[sinha10]

19 1

20 pan twist

21 ncameras

22 Levenberg-Marquardt GPS MRF

23 MRFBP

24 binary constraints: pairwise camera transformations & camera-point unary constraints: pose estimates (e.g., GPS, heading info) 3D points can also be modeled

25 matching all pairs of images 1) Vocabulary tree 2) GPS matched pairssiftann pairransac + 5 E

26 high-twist Rij relative twistrelative twist20 aspect ratios Internet MRF

27 p Camera 1 Camera 2 Camera 3 track3d

28 NodesMRFTracks 1camera-camera 2

29 MRFLabelingNP-hard grid-structured graphssfm MRFLabels Riti6Labels

30 6-dimensional label Rt twist 0 2D BP distance transformsbp messages [1] [1]P. Felzenszwalb and D. Huttenlocher, Efficient belief propa-gation for early vision, IJCV, vol. 70, no. 1, pp , 2006.

32 1 2 3MRFBP

33 Message L

34 T NODELABEL

35 BP

36 tiltpantwistxyz

37 twist = 0 11*11*11 [] tilt&pan LABEL LABEL482R

38 binary constraints: pairwise camera transformations unary constraints: pose estimates (e.g., GPS, heading info)

39 R ij IJ : t ij

40 twist = 0R (tilt & pan)

41 R ij t ij IJ tilttwist0

42 IJ tilttwist0

44 MRFmessage

45 MRFmessage TLabel

46 BP non-linear least squares twistrodrigues parameters

47 1 2 3BP

48 3D <x, y> BP LABELS2D LABELsz=0 GPS 300*300 GPS1~4

49 binary constraints: pairwise cameras Camera-3D point unary constraints: pose estimates (e.g., GPS, heading info) 3D points can also be modeled

50 R ij IJ : t ij

51 X Camera 1 Camera 3 Camera 2 Track3DX : Ray: XRay:

52 0

53 Cost as function of t j

54 R ij t ij IJ : GPS

56 Message 3DMessage

57 MRFCamerasmessage message

58 MRFCamerasmessage TLabel

59 BP non-linear least squares z

60 - 13D tracks3d 23Drobust Huber normb = 25

61 Reconstruction pipeline GPS 3D R, T P, X NLLS 2D2D 3D3D NLLS

62 MRFBP

63 3D 1IBA 22 3

65 Quad Total images: 6,514 Reconstructed images: 5,233 Edges in MRF: 995,734 Ground truth for 348 cameras Median error wrt ground truth: 1.16m (vs 1.01 for IBA)

67 Central Rome Reconstructed images: 14,754 Edges in MRF: 2,258,416 Median camera pose difference wrt IBA: 25.0m Our result Incremental Bundle Adjustment [Agarwal09]

69 GPS 100mBP

70 1 2 3LABEL 4BP

71 tilt0 Acropolistilt Quadtilt

72 3D Quad: 1.21m1.9m SanFrancisco: 4.94m4.97m Quad: 1.21m3.93m SanFrancisco: 4.94m7.14m

74 BP

75 1 SFM---MRFSFM Cameras3D 2IBA IBA6

76 LABELS - SFM

High Resolution Disparity Map Generation Using TOF Depth Camera In this paper, we propose a high-resolution disparity map generation method using a lo

High Resolution Disparity Map Generation Using TOF Depth Camera In this paper, we propose a high-resolution disparity map generation method using a low-resolution Time-Of- Flight (TOF) depth camera and

融合先验信息到三维重建 组会报 告[2]

融合先验信息到三维重建组会报告[2]