An Hausdorff distance based approach for evaluation of image moments in servoing applications

This paper presents a new criterion for the evaluation of visual features used in image based control laws. Due to their established performances in servoing applications, image moments derived from point features are evaluated. The proposed criterion analyses both the quantity and the quality of the information stored in an ensemble of point features used in controlling servoing systems. Using Hausdorff distance, new properties for point features ensemble are analyzed, for example the attached convex hull. In order to validate the proposed evaluation criterion, an environment composed from a six dof manipulator robot with an eye-in-hand configuration is used.