Adaptive Extended Kalman Filter for asynchronous shuttering error of stereo vision localization

Abstract

Stereo Vision (SV) Localization is an advanced sensory tool for motion estimation and tracking applications especially in robotics. Low cost and ease of use are the core benefits of this sensory system. One of the major problems in SV localization is the error in the estimation of object location because of the delay which is produced by asynchronous shuttering of each camera. A usual technique to overcome this problem is to make cameras operate synchronously. In this paper we propose Adaptive Extended Kalman Filter (AEKF) to compensate this asynchronization error by modeling the object's movement in image plane. In this approach after getting each cameras' frame we perform a localization process which results in a higher frame rate compared with the synchronous approach. It has also advantage of having more adaptable performance especially in the presence of uncertainty in arbitrary movements over conventional Kalman Filter and other linear techniques.