An Efficient Combination of 2D and 3D Shape Descriptions for Contour Based Tracking of Moving Objects

Tracking the 2D contour of a moving object has widely been used in the past years. So called active contour models have been proven to be a promising approach to real–time tracking of deformable objects. Also tracking 2D contours, which are projections of rigid 3D objects, is reduced to tracking deformable 2D contours. There, the deformations of the contour are caused by the movement in 3D and the changing perspective to the camera. In this paper a combination of 2D and 3D shape descriptions is presented, which can be applied to the prediction of changes in 2D contours, which are caused by movement in 3D. Only coarse 3D knowledge is provided, which is automatically acquired in a training step. Then, the reconstructed 3D model of the object is used to predict the shape of the 2D contour. Thus, limitations of the contour point search in the image is possible, which reduces the errors in the contour extraction caused by heterogenous background. The experimental part shows, that the proposed combination of 2D and 3D shape descriptions is efficient and accurate with respect to real–time contour extraction and tracking.