Lib4U

‎"Behind every stack of books there is a flood of knowledge."

Modeling Deformable Objects from a Single Depth Camera

f1

Miao Liao+ Qing Zhang+ Huamin WangRuigang YangMinglun Gong#
University of KentuckyGeorgia Institute of TechnologyMemorial University of Newfoundland#

 

We propose a novel approach to reconstruct complete 3D deformable models over time by a single depth camera, provided that most parts of the models are observed by the camera at least once. The core of this algorithm is based on the assumption that the deformation is continuous and predictable in a short temporal interval. While the camera can only capture part of a whole surface at any time instant, partial surfaces reconstructed from different times are assembled together to form a complete 3D surface for each time instant, even when the shape is under severe deformation. A mesh warping algorithm based on linear mesh deformation is used to align different partial surfaces. A volumetric method is then used to combine partial surfaces, fix missing holes, and smooth alignment errors. Our experiment shows that this approach is able to reconstruct visually plausible 3D surface deformation results with a single camera.

 

  

The inputs to our system are multiple color-depth image pairs captured at different time instants. In the initialization step, correspondences among salient features extracted from different frames are established by any tracking algorithms, typically SIFT in our experiment. In addition, the depth maps are triangulated into 3D meshes. The second step estimates a rigid transformation for each frame to map surfacepatches from all frames into a reference global coordinate, where they roughly align with each other. A linear mesh deformation method is then applied in the next step to warp one mesh to another, while preserving local details, which is the core of our algorithm. We break down the presentation into several subsections, first introducing the basics for pairwise warping, then extending it to a global alignment scheme. The issue of severe occlusion is also discussed. Finally in the smoothing and refinement step, meshes are merged into a single dynamic 3D model by volumetric methods, with temporal coherence among modelsfrom different frames enforced.

The real data is captured by a SwissRanger depth camera combined with a point grey flea color camera that provides texture information. The depth camera can produce 176 ?144 depth map at video rate. However, the quality of the depth map drops quite significantly for dynamic scenes. So we manually animate the toy giraffe and use temporal averaging to improve the signal-to-noise ratio of the depth map. SIFT features are extracted for each frame and features are tracked across different frames by searching in the pool of the extracted ones. The depth camera and color camera are almost coaxial so the mapping between their images can be approximated by a homography. There are about 200 features that are reliably tracked. Results in figure above show that the occlusion can be well handled by ouralgorithms. The above figure also shows the reconstruction of a T-shirt worn by a person who turned around 360 degrees in front of our capture device, while moving his hands up and down.

download:

Paper(pdf)

Video(AVI)

Source:

http://vis.uky.edu/~gravity/Research/4DRecWebPage/4DReconstruction.html

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Virtual Fashion Technology

Virtual Fashion Education

toitocuaanhem

"chúng tôi chỉ là tôi tớ của anh em, vì Đức Kitô" (2Cr 4,5b)

VentureBeat

News About Tech, Money and Innovation

digitalerr0r

Modern art using the GPU

Theme Showcase

Find the perfect theme for your blog.

lsuvietnam

Learn to Learn

Gocomay's Blog

Con tằm đến thác vẫn còn vương tơ

Toán cho Vật lý

Khoa Vật lý, Đại học Sư phạm Tp.HCM - ĐT :(08)-38352020 - 109

Maths 4 Physics & more...

Blog Toán Cao Cấp (M4Ps)

Bucket List Publications

Indulge- Travel, Adventure, & New Experiences

Lib4U

‎"Behind every stack of books there is a flood of knowledge."

The WordPress.com Blog

The latest news on WordPress.com and the WordPress community.

%d bloggers like this: