1 / 30

Abstract

Novel correspondence-based approach for consistent human skeleton extraction ( Springer US 2016). Abstract. A novel base-points-driven shape correspondence approach to extract skeletons of articulated objects from 3D mesh shapes. Introduction. Overview of this framework.

cassara
Download Presentation

Abstract

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. Novel correspondence-based approach for consistent human skeleton extraction (Springer US 2016)

  2. Abstract

  3. A novel base-points-driven shape correspondence approach to extract skeletons of articulated objects from 3D mesh shapes.

  4. Introduction

  5. Overview of this framework Topological thinning Mesh contraction Source shape Contracted mesh Joint-based skeleton offers Consistent information between the source shape and the target shapes BSC Finally, the skeletons of the target shapes of different poses are generated based on the correspondence relationship with source shape.

  6. Related work

  7. Consensus Skeleton for Non-rigid Space-time Registration (2010) • Curve skeleton extraction by coupled graph contraction and surface clustering (2013) • Kinematic skeleton extraction from 3D articulated models (2014) • Laplace-Beltrami spectra as Shape-DNA of surface and solids (2006) • Concise and provably informative multi-scale signature based on heat diffusion (2009)

  8. Base-points-driven shape correspondence(BSC)

  9. Downsampling methods Farthest Points Sampling(FPS) • Decimate the points to improve the efficiency of the algorithm • Algorithm Farthest Points Sampling(FPS): Sample_set= {} Randomly choose an initial point u i= 0 while i < n do choose a point v which is farthest from u add v into Sample_set i= i + 1 u = v

  10. Results of the FPS on two different shapes with 200 points

  11. New shape representation Consider the 3D surface as a 2D manifold surface: Step 1. Choose k base points Step 2. New coordinates can be expressed as: . . .

  12. Any point can be represented by the intersect point of three circles centered by three base points

  13. Heat Kernel and Heat Kernel Signature Considering the non-rigid deformations of these posture models, this makes it difficult to find their correspondence. So the HKS as the intrinsic features is chosen to find the base points.

  14. Heat Kernel Let be a compact Riemannian manifold without boundary, be a Laplace-Beltrami operator of The spectral expansion of heat kernel on any compact manifold has the following form:

  15. Heat Kernel Signature

  16. SIHKS(Scale-invariant Heat Kernel Signature) SIHKS is an extension of HKS. HKS SIHKS

  17. Base points are taken from the local maximum SIHKS in the range of a fixed geodesic distance on shapes.

  18. Examples of base points on different shape poses

  19. Identification of similar points Denote and as the source shape and the target shape respectively, where is the set of new points, which can be written as , is the number of base points. is the face from original shape. The goal is to find a mapping function:

  20. Correspondence results

  21. Skeleton extraction through shape correspondence

  22. Source shape skeleton extraction Curve Skeleton Extraction from Incomplete Point Cloud (2009) The suitable source shape skeleton has been extracted using above method. , where and are the constraints of contraction and attraction respectively.

  23. Consistent skeleton generation for target shape The closest shape vertex of source skeleton joints Where is minimum distance of the every skeleton joint from its associated shape vertex. Based on , the similar points of the target shape are grouped into one cluster.

  24. To connect the target shape nodes, a cyclic connection between nodes has been constructed based on shape correspondence and using rings of share vertices in target shape.

  25. Results and discussions

  26. Mesh contraction This method

  27. Conclusion

  28. This paper proposed an automatic and pose-invariant approach to extract the skeleton of an articulated human model based on BSC. • The skeleton generated by this approach does not require post-processing such as topological and geometric refinements.

  29. End

More Related