03-09-2012, 10:11 AM
ViBe:
ViBe.pptx (Size: 125.47 KB / Downloads: 52)
INTRODUCTION
Technique for motion detection.
Distinguish foreground from background.
Distinguish moving objects from static one.
REVIEW
Comparison of an observed images.
Adaptation to fast illumination changes.
Operates on each pixel independently.
Two level classifier mechanism .
Algorithm based on compressive sensing.
DESCRIPTION
Three consideration:
What is model and how does it behave?
How is the model initialized?
How is the model updated over time?
PIXEL MODEL AND CLASSIFICATION
Model each background with a set of samples.
Comparison is performed with the closest set of samples from a collection of samples.
Background pixel samples should be chosen carefully.
Classification of pixels should be done so that background pixels lie in the background only.
BACKGROUND MODEL INITIALIZATION
Uninterrupted foreground detection.
Response to sudden illumination changes.
Spatial neighborhood of each pixel.
Presence of ghosts