In this paper, we propose an approach for the control and layering of space-time continuous visual tasks with an uncalibrated camera. The approach is based on the bidimensional appearance of the objects in the environment, it allows the design of both reactive (or re exive) and active (purposive) tasks, and takes explicitly into account independent object motions. A linear model of camera-object interaction is embedded in the control scheme, which dramatically simpli es visual analysis and control by reducing the size of visual representation. The implementation of three visual tasks of increasing complexity, obtained with the proposed scheme and based on active contour analysis and polynomial planning of image contour transformations, is described and discussed. Both simulations and realtime experiments with a robotic eye-in-hand con guration are shown, which demonstrate the feasibility of the approach for applications in the elds of visual navigation, active exploration and percepti...
Carlo Colombo, James L. Crowley