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.


Related in: MedlinePlus

(a) The “T-pose” template model for the SMMC-10 dataset; (b) The labeled template; (c) The initial pose from a depth image; (d) The subject-specific articulated model obtained by transforming the template model.
© Copyright Policy
Related In: Results  -  Collection

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

f18-sensors-15-15218: (a) The “T-pose” template model for the SMMC-10 dataset; (b) The labeled template; (c) The initial pose from a depth image; (d) The subject-specific articulated model obtained by transforming the template model.

Mentions: The SMMC-10 dataset contains 28 depth image sequences (numbered 0 to 27) from the same subject with different motion activities, and it provides the ground-truth marker locations. The input depth image cannot be used directly, due to noise/outliers and undesirable background objects. Therefore, we performed three pre-processing steps to make the depth data ready for pose estimation, which include body subtraction by depth thresholding, a modified locally optimal projection (LOP) algorithm for denoising [22] and outlier removal by limiting the maximum allowable distance between two nearest points. Figure 17 shows an example of depth pre-processing for the SMMC-10 dataset. The “T-pose” template (around 2000 points) in this experiment is from [22], which has a built-in skeleton (Figure 18a) along with labeled body segments (Figure 18b). We selected one depth image with “T-pose” from Sequence 6 for shape initialization, which is given in Figure 18c, and the learned subject-specific shape model with a baked-in skeleton and labeled segments is shown in Figure 18d.


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 for the SMMC-10 dataset; (b) The labeled template; (c) The initial pose from a depth image; (d) The subject-specific articulated model obtained by transforming the template model.
© Copyright Policy
Related In: Results  -  Collection

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

f18-sensors-15-15218: (a) The “T-pose” template model for the SMMC-10 dataset; (b) The labeled template; (c) The initial pose from a depth image; (d) The subject-specific articulated model obtained by transforming the template model.
Mentions: The SMMC-10 dataset contains 28 depth image sequences (numbered 0 to 27) from the same subject with different motion activities, and it provides the ground-truth marker locations. The input depth image cannot be used directly, due to noise/outliers and undesirable background objects. Therefore, we performed three pre-processing steps to make the depth data ready for pose estimation, which include body subtraction by depth thresholding, a modified locally optimal projection (LOP) algorithm for denoising [22] and outlier removal by limiting the maximum allowable distance between two nearest points. Figure 17 shows an example of depth pre-processing for the SMMC-10 dataset. The “T-pose” template (around 2000 points) in this experiment is from [22], which has a built-in skeleton (Figure 18a) along with labeled body segments (Figure 18b). We selected one depth image with “T-pose” from Sequence 6 for shape initialization, which is given in Figure 18c, and the learned subject-specific shape model with a baked-in skeleton and labeled segments is shown in Figure 18d.

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.


Related in: MedlinePlus