Control of camera motions from the planning of image contours

A robotic system for camera positioning with respect to fixed planar objects is proposed, whose control is based on planning shape transformations of the objects' silhouette in the image. The control equations are linearized by means of an affine camera projection model, and even more simplified by maintaining fixation onto the objects during camera motions. No accurate camera calibration is required, and image contour-based visual analysis provides the system with a robust sensory feedback. The implementation with an eye-in-hand robotic setup is described and discussed. Applications are in the domains of purposive navigation and vision-based grasping.