Robust model-based motion tracking through the integration of search and estimation

Abstract

A computer vision system has been developed for real-time motion tracking of 3-D objects, including those with variable internal parameters. This system provides for the integrated treatment of matching and measurement errors that arise during motion tracking. These two sources of error have very dierent distributions and are best handled by separate computational mechanisms. These errors can be treated in an integrated way by using the computation of variance in predicted feature measurements to determine the probability of correctness for each potential matching feature. In return, a best-rst search procedure uses these probabilities to nd consistent sets of matches, which eliminates the need to treat outliers during the analysis of measurement errors. The most reliable initial matches are used to reduce the parameter variance on further iterations, minimizing the amount of search required for matching more ambiguous features. These methods allow for much larger frame-to-frame motions than most previous approaches. The resulting system can robustly track models with many degrees of freedom while running on relatively inexpensive hardware. These same techniques can be used to speed verication during model-based recognition.

Real-Time Tracking of Single and Multiple Objects from Depth-Colour Imagery Using 3D Signed Distance Functions

“…While limited computing power forced early authors (e.g. Harris and Stennett 1990;Gennery 1992;Lowe 1992) to exploit highly sparse data such as points and edges, the use of dense data is now routine.…”

A 2D–3D model-based approach to real-time visual tracking

“…A number of methods to compute pose from points have been proposed. We have used the method designed by Dementhon and Davis [8] completed by Lowe's non-linear method [22]. Using Dementhon's method, we calculate the rigid transformation iteratively using the coordinates of at least four points in the object coordinate system, and of their corresponding projections in the image as computed by the estimated 2D transformation: Ψ b Θ (X t ).…”

“…The latter explicitly exploits a model of the tracked objects. This model can be a 2D template of the object (active contours [2,18] or statistical modal representation [6,19]) or a CAD model [7,9,12,20,22,27,29]. CAD-based tracking methods attempt to register a reference model with the projection in the image of the desired object.…”

“…CAD-based tracking methods attempt to register a reference model with the projection in the image of the desired object. These approaches usually rely on a pose computation algorithm, involving the estimation of the 3D rigid transformation that links the object coordinate system to the camera coordinate system, from one view [22], multiple views or from an image sequence [7,9,12,20,29]. This class of methods usually obtains robust solutions (for example, it can cope with partial occlusion of the objects).…”

An efficient iterative pose estimation algorithm

“…where A i is a 3 by 3 matrix given by A i = I ; i t i : (15) For the derivation of the above formula, please refer to Appendix I.…”