Automatic Re-initialization of Head Pose Estimation
نویسندگان
چکیده
SYNOPSIS: The Vhuman-Embodiment project is a training simulation developed at the Institute of Creative Technologies in Los Angeles. In order to make the simulation more realistic, the user's head pose is used to provide the simulation with information on the user's attention. The purpose of this thesis has been to develop a system capable of estimating the head pose of the user. The system must be able to recover from failure, thus allowing it to run on for several minutes. The environment and the setup at the ICT-lab sets a number of limitations and restraints with regards to lighting and equipment for the head pose system. The solution is based on a model-based cylinder tracker, implemented at USC. A face detector is used to detect full frontal faces, at which point the cylinder tracker is re-initialized. Furthermore , the boundaries of the head are determined , and used for scaling of the cylinder during re-initialization. Tests show an overall average inaccuracy of 58 • , with a relative accuracy of 10.6 •. Further development is however required before the cylinder tracker is ready to be integrated with the system at ICT. This thesis implies that the reader holds the basic knowledge of real-time problems and fundamental knowledge related to image processing and computer vision. The sources for this report are indicated by [X], and are used for paragraphs or sentences. If it is given before a full stop, it means that it is only for the sentence, while after the full stop means that the source is for the whole paragraph. A DVD-ROM is included in the report, which includes the following: • An electronic version of this thesis in pdf format. • Test results, i.e. videos, of each of the system iterations as well as the overall system test in MPEG-V2 format. This thesis could not have been made without the help, guidance and support from the following group of people, which we would like to send a special thanks: Dr. Thomas Moeslund For being our supervisor during our final project at Aalborg University. And for initiating the contact with USC, allowing us to have a once in a lifetime experience abroad. For accepting us as visitors at USC and allowing us to be a part of the Vison group. And for taking over the job as overall project manager during our stay when it was needed. Postdoctoral Research …
منابع مشابه
Head pose estimation without manual initialization
This paper describes a fully automatic system for recovering the rigid components of head pose. It follows the conventional approach of tracking pose changes relative to a reference configuration. Rather than being supplied manually, here the reference configuration is determined by directly estimating pose from the image when the head is close to a frontal presentation. Tracking of pose is don...
متن کاملTowards Multilevel Human Body Modeling and Tracking in 3D: Investigation in Laplacian Eigenspace (LE) Based Initialization and Kinematically Constrained Gaussian Mixture Modeling (KC-GMM)
Vision-based automatic human body pose estimation has many potential applications and it is also a challenging task. Together, these two factors have made vision-based human body pose estimation an attractive research area with closely related research areas including body pose, hand pose, and head pose estimation. Up to now, these research works however only deal with each task of estimating b...
متن کاملتخمین چنددوربینی حالت سه بعدی انسان با برازش افکنش مدل اسکلت سه بعدی مفصل دار در تصاویر سایه نما
Automatic capture and analysis of human motion, based on images or video is important issue in computer vision due to the vast number of applications in animation, surveillance, biomechanics, Human Computer Interaction, entertainment and game industry. In these applications, it is clear that 3D human pose estimation is an essential part. Therefore, its accuracy has a great effect on the perform...
متن کاملHead Pose Estimation System Based on Particle Filtering with Adaptive Diffusion Control
In this paper, we propose a new tracking system based on a stochastic filtering framework for reliably estimating the 3D pose of a user’s head in real-time. Our system estimates the pose of a user’s head in each image frame whose 3D model is automatically obtained at an initialization step. In particular, our estimation method is designed to control the diffusion factor of a motion model adapti...
متن کاملHierarchical Approach for Articulated 3D Pose-Estimation and Tracking
In the recent years we presented a number of methods for a fully automatic pose estimation [5, 7] and tracking [6] of human bodies in 2D [5] and 3D [6]. Initialization and failure recovery in these methods are facilitated by the use of loose-limbed body model [7] in which limbs are connected via learned probabilistic constraints. The pose estimation and tracking can then be formulated as an inf...
متن کامل