A method to estimate the underlying motion and the structure of an unknown rigid object is presented. This approach differs from previous techniques in the relationship assumed between successive motions. With this assumption, the number of feature points can be reduced to only two which would otherwise be not interpretable by previous method[4] even if finitely many frames could be used. A closly t:e1ated assumption can be found in [1]. Several results are established: (1) There are two ambiguities when I.hree frames ace used. (2) There is a unique interpretation for !he underlying motion and structure when a fourth frame is included. (3) The uniqueness theorem was proved by a geomelric method which leads to an efficient, simple, and reliable algorithm. (4) Necessary crileria to rejectlhe suitabiltiy of the assumption are stated for three and four frames. (5) Many test examples are reported with detailed implementation experience. They reveal that the algorithm is stable.
[1]
Thomas S. Huang,et al.
Uniqueness and Estimation of Three-Dimensional Motion Parameters of Rigid Objects with Curved Surfaces
,
1984,
IEEE Transactions on Pattern Analysis and Machine Intelligence.
[2]
Berthold K. P. Horn,et al.
Determining Optical Flow
,
1981,
Other Conferences.
[3]
Jake K. Aggarwal,et al.
Structure from Motion of Rigid and Jointed Objects
,
1981,
Artif. Intell..
[4]
David F. Rogers,et al.
Mathematical elements for computer graphics
,
1976
.
[5]
I. Faux,et al.
Computational Geometry for Design and Manufacture
,
1979
.
[6]
Sudhir P. Mudur,et al.
Mathematical Elements for Computer Graphics
,
1985,
Advances in Computer Graphics.