Visual Servoing Using Relevant 2-D Image Cues

Abstract

In this paper, the authors demonstrate how some visual servoing behaviors can be achieved using direct 2-D image cues (both spatial and temporal). These cues serve as inputs to the motion control algorithms. In such cases, an explicit reconstruction of the scene is not required. The advantages of this 2-D based approach are that it is simple, fast and robust. There is no need for 3-D reconstruction. It results in a tight perception-action loop to directly generate action commands from image cues.