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.


Results of correspondence refinement before (above) and after (below) SAICP, especially around limb joints (circled area).
© Copyright Policy
Related In: Results  -  Collection

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

f13-sensors-15-15218: Results of correspondence refinement before (above) and after (below) SAICP, especially around limb joints (circled area).

Mentions: We first show some qualitative results of GLTP (α = 10, β = 2, λ = 5 × 106 and K = 10) by comparing with CPD in Figure 11 in terms of segment labeling accuracy. When articulated deformation is not significant between the template and target, such as the first pose, both CPD and GLTP perform well. However, in the cases of highly articulated deformations, e.g., Poses 2 to 5, significant labeling errors are observed around the head, limbs and body joints in the CPD results. On the other hand, GLTP provides stable segment label estimation across all poses. However, the results around limb joints are still not very reliable. We further perform the comparative analysis (averaged over 38 poses) with CPD, GLTP and AICP [16] in Figure 12, which shows that GLTP is the best one among all three, and AICP is better than CPD due to the fact that its locally rigid assumption is suitable for 3D human data. Figure 12 shows the labeling accuracy of body segments of our approach (GLTP + SAICP). It is shown that a significant improvement is achieved by using GLTP and SAICP jointly (GLTP + SAICP), which is also better than the one using CPD and SAICP together (CPD + SAICP). We visualize some labeling refinement results in Figure 13, where obvious improvements are seen around limb joints.


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

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

Results of correspondence refinement before (above) and after (below) SAICP, especially around limb joints (circled area).
© Copyright Policy
Related In: Results  -  Collection

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

f13-sensors-15-15218: Results of correspondence refinement before (above) and after (below) SAICP, especially around limb joints (circled area).
Mentions: We first show some qualitative results of GLTP (α = 10, β = 2, λ = 5 × 106 and K = 10) by comparing with CPD in Figure 11 in terms of segment labeling accuracy. When articulated deformation is not significant between the template and target, such as the first pose, both CPD and GLTP perform well. However, in the cases of highly articulated deformations, e.g., Poses 2 to 5, significant labeling errors are observed around the head, limbs and body joints in the CPD results. On the other hand, GLTP provides stable segment label estimation across all poses. However, the results around limb joints are still not very reliable. We further perform the comparative analysis (averaged over 38 poses) with CPD, GLTP and AICP [16] in Figure 12, which shows that GLTP is the best one among all three, and AICP is better than CPD due to the fact that its locally rigid assumption is suitable for 3D human data. Figure 12 shows the labeling accuracy of body segments of our approach (GLTP + SAICP). It is shown that a significant improvement is achieved by using GLTP and SAICP jointly (GLTP + SAICP), which is also better than the one using CPD and SAICP together (CPD + SAICP). We visualize some labeling refinement results in Figure 13, where obvious improvements are seen around limb joints.

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.