Limits...
Articulated Non-Rigid Point Set Registration for Human Pose Estimation from 3D Sensors.

Ge S, Fan G - Sensors (Basel) (2015)

Bottom Line: We introduce a visible point extraction method to initialize a new template for the current frame from the previous frame, which effectively reduces the ambiguity and uncertainty during registration.Third, to support robust and stable pose tracking, we develop a segment volume validation technique to detect tracking failures and to re-initialize pose registration if needed.The experimental results on both benchmark 3D laser scan and depth datasets demonstrate the effectiveness of the proposed framework when compared with state-of-the-art algorithms.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical and Computer Engineering, Oklahoma State University, Stillwater, OK 74078, USA. song.ge@okstate.edu.

ABSTRACT
We propose a generative framework for 3D human pose estimation that is able to operate on both individual point sets and sequential depth data. We formulate human pose estimation as a point set registration problem, where we propose three new approaches to address several major technical challenges in this research. First, we integrate two registration techniques that have a complementary nature to cope with non-rigid and articulated deformations of the human body under a variety of poses. This unique combination allows us to handle point sets of complex body motion and large pose variation without any initial conditions, as required by most existing approaches. Second, we introduce an efficient pose tracking strategy to deal with sequential depth data, where the major challenge is the incomplete data due to self-occlusions and view changes. We introduce a visible point extraction method to initialize a new template for the current frame from the previous frame, which effectively reduces the ambiguity and uncertainty during registration. Third, to support robust and stable pose tracking, we develop a segment volume validation technique to detect tracking failures and to re-initialize pose registration if needed. The experimental results on both benchmark 3D laser scan and depth datasets demonstrate the effectiveness of the proposed framework when compared with state-of-the-art algorithms.

No MeSH data available.


(a) The “T-pose” template model used for the SCAPE dataset; (b) The labeled template; (c) The labeled initial pose; (d) The learned subject-specific articulated model for SCAPE data (the estimated skeleton in black and the ground-truth one in blue).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4541828&req=5

f9-sensors-15-15218: (a) The “T-pose” template model used for the SCAPE dataset; (b) The labeled template; (c) The labeled initial pose; (d) The learned subject-specific articulated model for SCAPE data (the estimated skeleton in black and the ground-truth one in blue).

Mentions: The “T-pose” template used for the SCAPE data is modified from the MotionBuilder humanoid model, which has a skeleton and labeled body segments, as shown in Figure 9a, b, respectively. Given an initial pose from the SCAPE data that is close to the “T-pose”, we use the two-step approach discussed in Section 3.1 for shape initialization. Then, we obtain labeled body segments in Figure 9c and the estimated skeleton (joint positions) in Figure 9d. Compared with the ground-truth skeleton, the average error of joint positions is 2.88 cm. The subject-specific shape model shown in Figure 9d will be used in the following two experiments regarding correspondence estimation and pose estimation.


Articulated Non-Rigid Point Set Registration for Human Pose Estimation from 3D Sensors.

Ge S, Fan G - Sensors (Basel) (2015)

(a) The “T-pose” template model used for the SCAPE dataset; (b) The labeled template; (c) The labeled initial pose; (d) The learned subject-specific articulated model for SCAPE data (the estimated skeleton in black and the ground-truth one in blue).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4541828&req=5

f9-sensors-15-15218: (a) The “T-pose” template model used for the SCAPE dataset; (b) The labeled template; (c) The labeled initial pose; (d) The learned subject-specific articulated model for SCAPE data (the estimated skeleton in black and the ground-truth one in blue).
Mentions: The “T-pose” template used for the SCAPE data is modified from the MotionBuilder humanoid model, which has a skeleton and labeled body segments, as shown in Figure 9a, b, respectively. Given an initial pose from the SCAPE data that is close to the “T-pose”, we use the two-step approach discussed in Section 3.1 for shape initialization. Then, we obtain labeled body segments in Figure 9c and the estimated skeleton (joint positions) in Figure 9d. Compared with the ground-truth skeleton, the average error of joint positions is 2.88 cm. The subject-specific shape model shown in Figure 9d will be used in the following two experiments regarding correspondence estimation and pose estimation.

Bottom Line: We introduce a visible point extraction method to initialize a new template for the current frame from the previous frame, which effectively reduces the ambiguity and uncertainty during registration.Third, to support robust and stable pose tracking, we develop a segment volume validation technique to detect tracking failures and to re-initialize pose registration if needed.The experimental results on both benchmark 3D laser scan and depth datasets demonstrate the effectiveness of the proposed framework when compared with state-of-the-art algorithms.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical and Computer Engineering, Oklahoma State University, Stillwater, OK 74078, USA. song.ge@okstate.edu.

ABSTRACT
We propose a generative framework for 3D human pose estimation that is able to operate on both individual point sets and sequential depth data. We formulate human pose estimation as a point set registration problem, where we propose three new approaches to address several major technical challenges in this research. First, we integrate two registration techniques that have a complementary nature to cope with non-rigid and articulated deformations of the human body under a variety of poses. This unique combination allows us to handle point sets of complex body motion and large pose variation without any initial conditions, as required by most existing approaches. Second, we introduce an efficient pose tracking strategy to deal with sequential depth data, where the major challenge is the incomplete data due to self-occlusions and view changes. We introduce a visible point extraction method to initialize a new template for the current frame from the previous frame, which effectively reduces the ambiguity and uncertainty during registration. Third, to support robust and stable pose tracking, we develop a segment volume validation technique to detect tracking failures and to re-initialize pose registration if needed. The experimental results on both benchmark 3D laser scan and depth datasets demonstrate the effectiveness of the proposed framework when compared with state-of-the-art algorithms.

No MeSH data available.