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.


Some challenging cases in the SCAPE data (left: body segment labeling by GLTP; right: pose estimation by SAICP). The left arm/hand (a,b) and the right foot/leg (c) are mislabeled, which can be corrected during pose estimation. The two legs and feet (d,f) and the two hands and head (e) are labeled wrongly, which can be partially corrected by pose estimation.
© Copyright Policy
Related In: Results  -  Collection

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

f16-sensors-15-15218: Some challenging cases in the SCAPE data (left: body segment labeling by GLTP; right: pose estimation by SAICP). The left arm/hand (a,b) and the right foot/leg (c) are mislabeled, which can be corrected during pose estimation. The two legs and feet (d,f) and the two hands and head (e) are labeled wrongly, which can be partially corrected by pose estimation.

Mentions: The GLTP registration algorithm, which initializes the correspondences for SAICP-based articulated pose estimation, plays a critical role in the whole flow. Since GLTP uses the Euclidean distance to assign correspondences, it may not be reliable or valid in two challenging cases. First, when there is a strong pose articulation in the point set compared with the standard “T-pose” template, the EM-based GLTP optimization could be trapped into local minima, resulting in some body segments being wrongly labeled, which might be corrected by SAICP during pose estimation. Second, when some body segments are too close (the head and hands) or even merged (lower/upper legs), the shortest distance is no longer valid in those segments, leading to wrong correspondence estimation, which can only be partially corrected by SAICP due to large labeling errors. We further show six challenging cases in Figure 16, where the first row shows three examples of the first case and the second row presents three examples of the second case.


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

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

Some challenging cases in the SCAPE data (left: body segment labeling by GLTP; right: pose estimation by SAICP). The left arm/hand (a,b) and the right foot/leg (c) are mislabeled, which can be corrected during pose estimation. The two legs and feet (d,f) and the two hands and head (e) are labeled wrongly, which can be partially corrected by pose estimation.
© Copyright Policy
Related In: Results  -  Collection

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

f16-sensors-15-15218: Some challenging cases in the SCAPE data (left: body segment labeling by GLTP; right: pose estimation by SAICP). The left arm/hand (a,b) and the right foot/leg (c) are mislabeled, which can be corrected during pose estimation. The two legs and feet (d,f) and the two hands and head (e) are labeled wrongly, which can be partially corrected by pose estimation.
Mentions: The GLTP registration algorithm, which initializes the correspondences for SAICP-based articulated pose estimation, plays a critical role in the whole flow. Since GLTP uses the Euclidean distance to assign correspondences, it may not be reliable or valid in two challenging cases. First, when there is a strong pose articulation in the point set compared with the standard “T-pose” template, the EM-based GLTP optimization could be trapped into local minima, resulting in some body segments being wrongly labeled, which might be corrected by SAICP during pose estimation. Second, when some body segments are too close (the head and hands) or even merged (lower/upper legs), the shortest distance is no longer valid in those segments, leading to wrong correspondence estimation, which can only be partially corrected by SAICP due to large labeling errors. We further show six challenging cases in Figure 16, where the first row shows three examples of the first case and the second row presents three examples of the second case.

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.