Abstract. A computer vision software library is a key component of vision-based applications. While there are several existing libraries, most are large and comple or limited to a particular hardware/platform combination. These factors tend to impede the d
TLIB: a Real-time Computer Vision Library for HCI
Sébastien Grange, Terrence Fong, and Charles Baur
Virtual Reality and Active Interfaces (VRAI) Group
Institut de production et robotique, Swiss Federal Institute of Technology (EPFL)
CH-1015 Lausanne, Switzerland
{sebastien.grange, terrence.fong, charles.baur}@epfl.ch
Abstract. A computer vision software library is a key component of vision-based
applications. While there are several existing libraries, most are large and comple or
limited to a particular hardware/platform combination. These factors tend to impede
the development of research applications, especially for non-computer vision experts.
To address this issue, we have developed TLIB, an easy-to-learn, easy-to-use software
library that provides a complete set of real-time computer vision functions, including
image acquisition, 2D/3D image processing, and visualization. In this paper, we
present the motivation for TLIB and its design. We then summarize some of the
applications that have been developed with TLIB, and discuss directions for future
work.
1. Introduction
Since 1999, the EPFL Virtual Reality and Active Interfaces (VRAI) Group has been
developing non-traditional human-computer interfaces in a variety of fields, including
computer assisted surgery and mobile robotics. These interfaces exploit numerous
interaction techniques based on Computer Vision (CV), such as activity monitoring, human
detection and tracking, and gesture recognition. The VRAI group, for example, is part of
the CO-ME network [2], which focuses on the application of information technology to
medical treatment. One area of interest is increasing the use of computer equipment in the
operating room (OR). Because OR’s are crowded environments and have stringent sterility
requirements, traditional computer input devices (i.e., keyboard and mouse) are
problematic. Thus, there is a significant need to develop non-contact (i.e., vision-based)
interaction methods.
Another area we have been investigating is teleoperation interfaces for mobile robots.
Traditional remote driving systems are based on hand-controllers and video displays. Such
systems are error-prone, time consuming to deploy, and difficult to use when joint (i.e.,
human-robot) task performance is required. An alternative is to enable the human to
interact directly with the robot, e.g., communicating commands through hand gestures.
Autonomous robots however, typically have limited processing power. Hence, the vision
system must be efficient and lightweight.
Finally, many of our research projects involve designing and demonstrating application
prototypes. Very often, undergraduate students with limited programming and computer
vision experience contribute to such developments as part of their degree program. For
these students to be productive, it is important to minimize the overhead associated with
learning and implementing fundamental computer vision methods, such as camera
calibration and pixel operators.
1017
vision software library is a key component of vision-based applications. While...2003, Sydney TLIB: a Real-time Computer Vision Library for HCI Sébastien...
热门文档