Range identification of a mobile robot with a monocular camera system

Range identification is a classical problem in the field of machine vision, which aims to recover the depth information of a moving object/camera from 2D image space for perspective dynamic systems (PDS). In this paper, range identification for a specific constrained vision system-a mobile robot with a monocular camera (MRMC) is discussed. First, the dynamics of the perspective vision system is modeled with the range information as an unknown state variable, leading to a nonlinear system in which 2D image coordinates of the reference point are measurable. Then, inspired by the recent reduced-order observer (ROO) design technique, we apply the ROO method directly for the specific MRMC system with full observation. Furthermore, a ROO-based nonlinear observer with a single observation is also proposed to cope with the situation when full observation cannot be obtained. Finally, the performance of the proposed observers are demonstrated through simulation results in which image noises are considered to be more close to practical systems.