Real-time 3D object tracking

Abstract A system for tracking the three dimensional motion of objects in realtime is described. The system uses a technique of localized image access that enables it to achieve frame—rate operation without the need for any special-purpose hardware. All the processing for one camera can be performed on a network of five transputers, with a single transputer combining the information from several cameras. A dynamic Hough transform is used to solve a set of constraints derived from image measurements. It is demonstrated how the robustness and intrinsic parallelism of the Hough transform can be put to practical use. The system has been tested on a stored sequence of 100 image pairs, which were captured under realistic, noisy conditions, and the results are presented.

[1]  Hungwen Li,et al.  Fast Hough transform: A hierarchical approach , 1986, Comput. Vis. Graph. Image Process..

[2]  Thomas S. Huang,et al.  Estimation of rigid body motion using straight line correspondences , 1986, Comput. Vis. Graph. Image Process..

[3]  Daram V. Ahuja,et al.  Geometry for Construction and Display , 1968, IBM Syst. J..

[4]  John F. Canny,et al.  A Computational Approach to Edge Detection , 1986, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[5]  Kay Brisdon,et al.  Alvey MMI-007 Vehicle Exemplar: Evaluation and Verification of Model Instances , 1987, Alvey Vision Conference.

[6]  D. W. Thompson,et al.  Three-dimensional model matching from an unconstrained viewpoint , 1987, Proceedings. 1987 IEEE International Conference on Robotics and Automation.