Adaptive model-based motion estimation

Abstract

A general discrete-time, adaptive, multidimensional framework is introduced for estimating the motion of one or several object features from their successive non-linear projections on an image plane. The motion model consists of a set of linear difference equations with parameters estimated recursively from a non-linear observation equation. The model dimensionality corresponds to that of the original, non-projected motion space thus allowing to compensate for variable projection characteristics such as panning and zooming of the camera. Extended recursive least-squares and linear-quadratic tracking algorithms are used to adaptively adjust the model parameters and minimize the errors of either smoothing, filtering or predicting the object trajectories in the projection plane. Both algorithms are derived using a second order approximation of the projection nonlinearities. All the results presented here use a generalized vectorial notation suitable for motion estimation of any finite number of object features and various approximations of the nonlinear projection. An example of motion estimation in a sequence of video frames is given to illustrate the main concepts.