Change and Motion Estimation
In this chapter, we examine techniques to estimate motion and change from a sequence of spatial grids when what is being observed is moving as well as evolving. We consider first simply subtracting successive grids and point out the limitations of this approach. Then, we consider using partial derivatives (optical flow) which is suitable for fluid-like flows. Cross-correlation is often better than using partial derivatives when the amount of change and movement are large. We examine a way to improve cross-correlation, by performing it in the frequency domain (phase correlation). Then, we discuss object tracking which is suitable when the spatial grid consists of objects that are moving and changing rather than of fluid-like flows. Object tracking involves associating objects between frames, and this leads us to a discussion of the Hungarian method. Finally, we point out that a hybrid approach allows us to retain the advantages of both object tracking and cross-correlation while side-stepping their disadvantages. We finish the chapter by discussing different ways of computing temporal attributes from spatial grids.
KeywordsOptical Flow Motion Vector Motion Estimate Object Tracking Current Frame
- Brown RG, Hwang P (1997) Introduction to random signals and applied Kalman filtering. Wiley, New YorkGoogle Scholar
- Michel R, Rignot E (1999) Flow of Moreno Glaciar, Argentina, from repeat-pass shuttle imaging radar images: comparison of the phase correlation method with radar interferometry. J Glaciol 45:93–100Google Scholar
- Rudlosky S (2010) Assessing storm severity using lightning and radar information. PhD thesis, Florida State UniversityGoogle Scholar