Modeling of Arbitrary-Shaped Objects by Evolving-Ellipsoids Approximation for Visual Servoing

Model-based visual servoing needs definitions of the objects and detected pose of the model is used for visual servo controller. When 3-D pose-based visual servoing is applied to 3-D unknown-shaped target, the model-based visual servoing method needs create the artificial 3-D model definitions for detecting the unknown 3-D target and tracking it. In order to have visual servoing system track the unknown-shaped target, automatic 3-D model generation mechanisms for arbitrary-shaped target objects should be created before detecting the designated target. In this paper, a 3-D modeling method of arbitrary shape objects is proposed and the performance is examined by modeling experiment of arbitrary shape objects and visual servoing experiments with twin hand-eye camera whose looking direction can rotate around pitching and yawing axes.