Single camera based motion and shape estimation using extended Kalman filtering (Q1600430)

The problem of estimating the motion and shape parameters of a moving rigid body from a time series of noisy image data is considered in the paper. In general, the estimation problem is ill-posed and is quite sensitive to noise in the observed data. In order to overcome the noise problem, various recursive estimation schemes were proposed. The most popular of which is the extended Kalman filtering procedure. The authors propose a numerical study and compare a recursive algorithm based on an extended Kalman filter with a non-recursive algebraic method of estimating the motion and shape parameters of a moving object. The conclusion of the paper is that the recursive methods give robust estimates while the algebraic estimation schemes are sensitive to noise.