Human Pose Estimation in Stereo Images

Abstract

In this paper, we address the problem of 3D human body pose estimation from depth images acquired by a stereo camera. Compared to the Kinect sensor, stereo cameras work outdoors having a much higher operational range, but produce noisier data. In order to deal with such data, we propose a framework for 3D human pose estimation that relies on random forests. The first contribution is a novel grid-based shape descriptor robust to noisy stereo data that can be used by any classifier. The second contribution is a two step classification procedure, first classifying the body orientation, then proceeding with determining the full 3D pose within this orientation cluster. To validate our method, we introduce a dataset recorded with a stereo camera synchronized with an optical motion capture system that provides ground truth human body poses.

Human Pose Estimation from Monocular Images: A Comprehensive Survey

“…Other features, such as local paths [231], prediction pipeline [232], and Extremal Human Curves [233] are also common features in human pose estimation.…”

Stereo Pictorial Structure for 2D articulated human pose estimation

“…Since the authors work on video sequences, the body pose estimated in the previous frame is used to initialize the current pose. Random forests are used in [31] to estimate the 3D body pose from stereo images in two steps. Firstly, a grid-based shape descriptor is computed from depth maps to predict the orientation of the body.…”

“…Although the authors present promising results on their custom dataset of ten people performing controlled movements, we cannot predict the behaviour of their method on uncontrolled scenarios, as the ones used in this paper. Note that neither [50] nor [31] make use of RGB data during the pose estimation process and, therefore, a good quality of the estimated depth maps is required to obtain accurate results, even in controlled indoor scenarios. In contrast, in this paper, the combination of both RGB and depth information allows us to deal with a wide range of body poses and very challenging imaging conditions.…”

A Methodology for Multi-Camera Surface-Shape Estimation of Deformable Unknown Objects

“…The methodology is based on the use of a multi-camera system, with a variable number of cameras, and range of object motions. The numerous simulations and experiments presented herein demonstrate the proposed methodology's ability to accurately estimate the surface deformation of unknown objects, as well as its robustness to object loss under self-occlusion, and varying motion dynamics.Typical motion capture methods utilize an articulated object model (i.e., a skeleton model) that is fit to the recovered 3D data [34][35][36][37][38]. The deformation of articulated objects is defined as the change in pose and orientation of the articulated links in the object.…”

“…Typical motion capture methods utilize an articulated object model (i.e., a skeleton model) that is fit to the recovered 3D data [34][35][36][37][38]. The deformation of articulated objects is defined as the change in pose and orientation of the articulated links in the object.…”