A geometric moment-based observer for global range and local orientation identification

A reduced-order globally convergent observer to estimate the depth of an object projected on the image plane of a camera is presented, assuming that the object is planar or has a planar limb surface and the orientation of the plane is known. A local result can be obtained when the plane unit normal is unknown, and the latter is estimated together with the depth of the object. The observer exploits the image moments of the object as measured features. The estimation is achieved by rendering attractive and invariant a manifold in the extended state space of the system and the observer. The problem is reduced to the solution of a system of partial differential equations.

I. Introduction

Vision guided robotics has been one of the major research issues in the last years, since the applications of visually guided systems are numerous, e.g. intelligent agents, robotic surgery, exploration rovers and home automation [6], [8]. The use of direct visual information, which is acquired from measurable characteristics of the environment, to provide feedback about the state of the environment itself and to control a robot is commonly termed Visual Servoing, These observed characteristics are usually referred to as features. Features are extracted from the image and their motion is mapped to the velocity twist of the camera via an interaction matrix [8]. The image of the target is a function of the relative pose between the camera and the target and the distance between them is frequently referred to as depth or range.