Rigid and affine motion estimation in vision using sliding mode observers

A "sliding mode observer" based approach to estimate rigid and affine motion of an object is presented in this paper. Two different problems are considered. First the motion is of rigid type and constrained to a plane, and object, e.g. a mobile robot, is viewed using a stationary camera at a fixed certain distance in the normal direction of the motion plane, thus the motion in image plane also being rigid. Then we consider the problem of identifying parameters of an affine motion which may result as the scaled orthographic projection of an almost planar object undergoing 3D rigid motion. In either case, certain number of feature points extracted from the images of the object is used for parameter estimation. Though we know these feature points in every frame, we also construct their estimates, and try to make these estimates follow the actual ones using sliding mode control. It turns out that the control employed in this framework allows us to estimate the motion parameters of the moving object in an elegant way.