Finding Articulated Body in Time-Series Volume Data

Abstract

Abstract. This paper presents a new scheme for acquiring 3D kinematic structure and motion from time-series volume data, in particular, focusing on human body. Our basic strategy is to first represent the shape structure of the target in each frame by using aMRG, augmented Multiresolution Reeb Graph [6], and then deform each of the shape structures so that all of them can be identified as a common kinematic structure throughout the input frames. Although the shape structures can be very different from frame to frame, we propose to derive a unique kinematic structure by way of clustering some nodes of graph, based on the fact that they are partly coherent. The only assumption we make is that human body can be approximated by an articulated body with certain number of end-points and branches. We demonstrate the efficacy of the proposed scheme through some experiments.

Local Equivalence and Intrinsic Metrics between Reeb Graphs

“…realized as the lengths of shortest continuous paths in the space of Reeb graphs [9]. This is particularly useful when one actually needs to interpolate between data, and not just discriminate between them, which happens in applications such as image or 3-d shape morphing, skeletonization, and matching [17,20,21,27]. At this time, it is unclear whether the metrics proposed so far for Reeb graphs are intrinsic or not.…”

Finding Kinematic Structure in Time Series Volume Data

“…Moreover, the scope of motion description is limited to the case that the precise structure of the target is given. For these reasons, vision-based human motion capture is an active research area these days [1,7]. In order to realize a scheme for describing arbitrary moving objects which can be approximated by articulated rigid body without using any special markers and suits, we employ time series volume data as the input which we compute from multiviewpoints videos.…”

Towards a model-based 3D marker-less human motion capture

“…Other works attempt to improve the skeleton estimation by labeling body parts to kinematics chain of super-quadrics representing voxel data [144]. Also in [98], they derive a common kinematic structure through representing the shape of the subject in each time frame by using augmented Multiresolution Reeb Graph [62]. However, we have to take note that the skeletons obtained mathematically are not the same as anatomic skeletons and do not take into consideration the skin deformations, hence numerical skeletons will not give the true anatomic joint locations e.g.…”