Abstract
A general problem in motion analysis when applied to motion prediction, trajectory planning, or autonomous vehicle navigation is the need for accurate determination of three-dimensional displacement parameters in a dynamic scene. Data are acquired by identifying point correspondences from a stereo pair of images obtained at two different time instances. The three-dimensional location of points on an object is found by performing a triangulation on each of the stereo image pairs and is fed to a motion-estimation algorithm. The uncertainty arising from the large amount of noise in this procedure at typical image resolutions and its effect on motion estimation algorithm performance are investigated. Issues in the design of robust motion estimation algorithms are also discussed. Simulation results that confirm the analysis and demonstrate algorithm performance are included.
| Original language | English (US) |
|---|---|
| Title of host publication | Unknown Host Publication Title |
| Publisher | IEEE |
| Pages | 518-523 |
| Number of pages | 6 |
| ISBN (Print) | 0818606339 |
| State | Published - 1985 |
ASJC Scopus subject areas
- Engineering(all)