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.


Comparative analysis of the joint estimation error (cm).
© Copyright Policy
Related In: Results  -  Collection

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

f20-sensors-15-15218: Comparative analysis of the joint estimation error (cm).

Mentions: Our proposed framework is also compared against some recent state-of-the-art algorithms [13,18–20,22,49,50] in terms of the error between each estimated joint and its corresponding ground-truth marker. Given a sequence with Nf frames and Nj joints, the joint estimation error is defined as:(13)e=1NfNj∑k=1Nf∑i=1Nj‖Jik−Mik−Oi‖where and are the estimated position and the ground-truth marker position of the i-th joint in the k-th frame. Due to the inconsistency between the definition of joints between the template skeleton and the configuration of markers, we need to remove a constant offset Oi at each joint that is computed along the local segment based on 20 manually-selected frames. Figure 18c,d show the initial pose from the depth image and the learned subject-specific shape model with labeled segments and the estimated skeleton, respectively. The quantitative comparison against several recent algorithms in terms of the position error (averaged over all frames from 28 sequences) is shown in Figure 20. The accuracy of pose estimation is significantly improved compared with that in [49] (4.3 cm) due to the tracking capability, including visible point extraction and segment volume validation. The average joint position error is 3.2 cm, which outperforms all existing methods, including the most recent work [18] (3.4 cm).


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

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

Comparative analysis of the joint estimation error (cm).
© Copyright Policy
Related In: Results  -  Collection

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

f20-sensors-15-15218: Comparative analysis of the joint estimation error (cm).
Mentions: Our proposed framework is also compared against some recent state-of-the-art algorithms [13,18–20,22,49,50] in terms of the error between each estimated joint and its corresponding ground-truth marker. Given a sequence with Nf frames and Nj joints, the joint estimation error is defined as:(13)e=1NfNj∑k=1Nf∑i=1Nj‖Jik−Mik−Oi‖where and are the estimated position and the ground-truth marker position of the i-th joint in the k-th frame. Due to the inconsistency between the definition of joints between the template skeleton and the configuration of markers, we need to remove a constant offset Oi at each joint that is computed along the local segment based on 20 manually-selected frames. Figure 18c,d show the initial pose from the depth image and the learned subject-specific shape model with labeled segments and the estimated skeleton, respectively. The quantitative comparison against several recent algorithms in terms of the position error (averaged over all frames from 28 sequences) is shown in Figure 20. The accuracy of pose estimation is significantly improved compared with that in [49] (4.3 cm) due to the tracking capability, including visible point extraction and segment volume validation. The average joint position error is 3.2 cm, which outperforms all existing methods, including the most recent work [18] (3.4 cm).

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.