Model-based stereo-visual tracking: Covariance analysis and tracking schemes

Abstract This paper addresses the problem of target tracking in stereo images from two perspectives. Initially, we carry out an eigenanalysis of the covariance matrix of the reconstruction error for a simple but realistic case. The analysis gives insight into the problem and it also leads naturally to the second issue addressed in the paper, namely, the analysis and design of variations of the well-known Extended Kalman filter. These variations are a trade-off between performance and computational complexity. A deep study of the conditions under which every alternative works correctly is developed. Our conclusions are supported by a number of Monte Carlo simulations using the stereo scheme proposed in the first part of the paper.