Computer methods for 3D motion tracking in real-time

Boguslaw Rymut; Computer methods for 3D motion tracking in real-time; PhD Thesis; 2016
Supervisor: Bogdan Kwolek, DSc, PhD, Eng., Associate Prof
Keywords: model-based 3D motion tracking, parallel computing, computer vision

Abstract

This thesis is devoted to marker-less 3D human motion tracking in calibrated and synchronized multicamera systems. Pose estimation is based on a 3D model, which is transformed into the image plane and then rendered. Owing to elaborated techniques the tracking of the full body has been achieved in real-time via dynamic optimization or dynamic Bayesian filtering. The objective function of a particle swarm optimization algorithm and the observation model of a particle filter are based on matching between the rendered 3D models in the required poses and image features representing the extracted person. In such an approach the main part of the computational overload is associated with the rendering of 3D models in hypothetical poses as well as determination of value of objective function. Effective methods for rendering of 3D models in real-time with support of OpenGL as well as parallel methods for determining the objective function on the GPU were developed. Several variants of objective function using both software and hardware rendering were proposed and evaluated on real data. Methods for effective rendering of 3D models in OpenGL, as well as data mapping between OpenGL and CUDA were developed and evaluated. Programmable streams in OpenGL were designed and configured to achieve real-time rendering of considerable number of 3D models in desired poses. Methods for parallel execution of particle swarm optimization as well as objective function calculation were developed to achieve effective utilization of hardware resources and the best possible tracking accuracies and frequencies. The elaborated solutions permit 3D tracking of full body motion in real-time. Certain solutions to enable selection of the number of particles and the number of iterations in the PSO algorithm, which determine the number of processed frames per second, and which in turn determines the change in the pose between consecutive frames were investigated and proposed. They make it possible to achieve the lowest errors in real-time 3D motion tracking.

Media

PhD Thesis in PDF format
Citation in BibTeX format

Citation


@PhdThesis{RymutPhdThesis2016, author = {Bogusław Rymut}, title = {Komputerowe algorytmy ekstrakcji i śledzenia obiektów w czasie rzeczywistym}, year = {2016}, }

Full Body Motion Tracking in Monocular Images Using Particle Swarm Optimization

Bogusław Rymut and Tomasz Krzeszowski and Bogdan Kwolek: Full Body Motion Tracking in Monocular Images Using Particle Swarm Optimization; Int. Conf. on Artificial Intelligence and Soft Computing, 2012, Artificial Intelligence and Soft Computing, Lecture Notes in Computer Science, Vol. 7267, Springer Berlin Heidelberg, 2012, pp. 600-607.

Abstract

The estimation of full body pose in monocular images is a very difficult problem. In 3D-model based motion tracking the challenges arise as at least one-third of degrees of freedom of the human pose that needs to be recovered is nearly unobservable in any given monocular image. In this paper, we deal with high dimensionality of the search space through estimating the pose in a hierarchical manner using Particle Swarm Optimization. Our method fits the projected body parts of an articulated model to detected body parts at color images with support of edge distance transform. The algorithm was evaluated quantitatively through the use of the motion capture data as ground truth.

Citation

@incollection{
    title={Full Body Motion Tracking in Monocular Images Using Particle Swarm Optimization},
    year={2012},
    isbn={978-3-642-29346-7},
    booktitle={Artificial Intelligence and Soft Computing},
    volume={7267},
    series={Lecture Notes in Computer Science},
    editor={Rutkowski, Leszek and Korytkowski, Marcin and Scherer, Rafał and Tadeusiewicz, Ryszard and Zadeh, LotfiA. and Zurada, JacekM.},
    doi={10.1007/978-3-642-29347-4_70},
    url={http://dx.doi.org/10.1007/978-3-642-29347-4_70},
    publisher={Springer Berlin Heidelberg},
    author={Rymut, Bogusław and Krzeszowski, Tomasz and Kwolek, Bogdan},
    pages={600-607}
}

GPU-Supported Object Tracking Using Adaptive Appearance Models and Particle Swarm Optimization

Boguslaw Rymut and Bogdan Kwolek: GPU-Supported Object Tracking Using Adaptive Appearance Models and Particle Swarm Optimization; Int. Conf. on Computer Vision and Graphics, 2010, Computer Vision and Graphics, Lecture Notes in Computer Science, Vol. 6375, Springer Berlin Heidelberg, 2010, pp. 227-234.

Abstract

In this paper we present a particle swarm optimization (PSO) based approach for marker-less full body motion tracking. The objective function is smoothed in an annealing scheme and then quantized. This allows us to extract a pool of candidate best particles. The algorithm selects a global best from such a pool to force the PSO jump out of stagnation. Experiments on 4-camera datasets demonstrate the robustness and accuracy of our method. The tracking is conducted on 2 PC nodes with multi-core CPUs, connected by 1 GigE. This makes our system capable of accurately recovering full body movements with 14 fps.

Citation