01-31-2011 08:20 AM
hello
I'm looking to use optical flow to measure the speed of an object (e.g.. a car). Segmenting the velocity image to look for the object works fine but I'm confused about the output units of the optical flow algorithms. It seems that the magnitude of the output is changed when you change e.g. the "Smoothing Parameter". Can someone explain this to me?
pieter
02-07-2011 08:10 AM
Hello,
In LabVIEW Help it is specified that "the Smoothing Parameter specifies the smoothing effect to apply to the velocity information in the image."
The Smoothing Parameter itself should be a positive value.
If you have a large (relative) motion between two image frames or between two video frames, then the smoothing parameter should also be large.
In case of a small motion between two frames, you should use a small smoothing parameter.
As you might think these small and large values are in most cases determined based on empirical data or on previous experiences of the operator.
If you're new to these kind of algorithms, then it would be useful to first try the included examples in LabVIEW (eg. Optical Flow Example) to get an idea of the effect of changing this parameter.
There are also several papers about this topic available online.
A google search with the terms "optical flow smoothing parameter" will result in several papers about this.
If this is not clear or there are more specific questions, then don't hesitate to ask me.
01-19-2018 01:43 PM
This is an extremely unsatisfying answer. Please provide the formula or any other detailed explanation what the "smoothing parameter" does.