Abstract:
In order to build a complete vision-based control system we need the
integration of several research areas as visual matching, visual
tracking and visual servoing. System integration is becoming an
important issue due to the high number of different methods for
tracking and servoing. This paper is the first step towards the
integration of several recent research works in computer vision and
robotics that authors have proposed in complementary research
areas. The integration is made in a unifying framework which allows
the use of model-free vision-based control techniques. These
techniques have been designed to control a robot with respect to
objects of any shape (i.e. without knowing a CAD model). Most systems
have been designed for positioning a robot by tracking specific visual
features (e.g. interest points, straight lines, contours). If such
features are not available, several parts of the system must be
designed again. In this paper, we propose a more flexible approach
which can be used even with textured objects with no evident
structured features. The main contribution of this paper is to
integrate into an unifying framework recent advances in real-time
template-based visual tracking and model-free visual servoing
techniques. |