Abstract: We propose an approach for the design and control of both reflexive and purposive visual tasks with an uncalibrated camera. The approach is based on the bi-dimensional appearance of the objects in the environment, and explicitly takes into account independent object motions. The introduction of a linear model of camera-object interaction dramatically simplifies visual analysis and control by reducing the size of the visual representation. We discuss the implementation of three tasks of increasing complexity, based on active contour analysis and polynomial planning of image contour transformations. Real-time experiments with a robot wrist-mounted camera demonstrate that the approach is conveniently usable for visual navigation, active exploration and perception, and man-robot interaction.
0 Replies
Loading