| Summary: | Thesis (MScEng)--University of Stellenbosch, 2002. === ENGLISH ABSTRACT: Structure from Motion is defined as the problem of extracting the 3d motion of a camera
moving through a scene, as well as the 3d structure of the scene, from the image sequence
produced by the camera over time. Several methods based on the Kalman filter have
been proposed in the past, mostly based on the Extended Kalman filter. We propose
an algorithm based on the dual Unscented Kalman filter to estimate the structure and
motion of an object under perspective projection. It is shown that the algorithm is stable
and accurate under synthetic as well as real-world conditions. === AFRIKAANSE OPSOMMING: Struktuur vanuit Beweging is 'n rekenaar-visie probleem waarin die 3d beweging van 'n
kamera deur 'n ruimte, asook die 3d struktuur van die ruimte, bepaal moet word slegs
vanuit die 2d beelde in die beeldreeks wat deur die kamera geneem word. 'n Verskeie reeks
oplossings, gebaseer op die Kalman filter, is reeds voorgestelom die probleem op te los.
Meeste van die oplossings implementeer die "Extended Kalman filter", of EKF. Ons stel
'n algoritme voor, gebaseer op 'n nuwe nie-lineêre benadering tot die Kalman filter, die
sogenaamde "Unscented Kalman filter", of UKF. Hierdie algoritme bepaal die struktuur
en beweging onder 'n perspektief-projeksie kamera. Daar word getoon dat die algoritme
stabiel en akkuraat funskioneer onder sintetiese sowel as reële toevoer.
|
|---|
