It is hard to find optical flow on featureless planar. Maybe you can try to add other sensors.

thanks for your advice to xf he.

Dear Gi Dong Kim,

If you are using a wheeled robot and a vision system, you can mark some signs on wheels. So you can calculate linear and angular velocity observing these signals on the wheels. You know the wheel's parameters and you can estimate linear velocity and angular velocity.

I hope that answer be helpful

I have one paper that a use a omnidiretional vision system. So I can look for all environment around the robot. I use some texture on objets and also I have acesses to robot encoders. But in your problem, you can use this system to watch robot wheels.

Best regards.

André S Franca

Article DETECÇÃO DE OBSTÁCULOS ATRAVÉS DE UM FLUXO ÓPTICO PADRÃO OBT...

Dear Gi Dong Kim,

If you *know* for sure that what you are looking at is a featureless plane, then you can compute the optical flow if you know the distance and orientation of your camera relative to the plane. You could get these by projecting a structured light pattern onto the plane (of course this requires prior calibration of your camera + structured light system).

But then, if you end up using structured light, this raises the question of why you want to get the optical flow at all. Generally, the optical flow is computed as a means to extract more "useful" information (e.g. depth or time to collision). Here you would be going the other way, starting from what is generally an "end result" (distance and orientation) to compute the optical flow.