Optical flow data

[siva karthik]

I have been trying to use the Centeye Sonyman chip for obstacle avoidance using the optical flow it gives in an environment. But I havent made any breakthrough except that I am getting the the optical flow reading in the serial port monitor.

It would be greatly helpful iff you could let me know on how that stream of number is is organized and how I need to interpret it so that I can use the data for obstacle avoidance. In all, I want to know how to use that optical flow data for obstacle avoidance. Please help me for which I would be grateful to you. 

[siva karthik]

In continuation to the earlier mail, I have studied in detail on how the optical flow algorithms evaluate optical flow using images. Also I have clearly read on Lucas Kanade method of evaluating optical flow as well.

What I want to know is that when I run the OFO sample code give in the xamples, I see that the optical flow is being estimated and displayed only as the centre of the image and not at all the points. Could you please give me insights on what this means and how I can use this for obstacle avoidance?

[Geoffrey L. Barrows]

Hi Siva,

The reason you see just one optical flow measurement displayed in the center of the image is that there is only one optical flow value being computed. If you want to see more vectors, you'll have to modify the code so that the image is divided into smaller sub-imges, with an optical flow value computed for each sub-image.

To answer your previous question about how to use optical flow to avoid objects- there are many possibilities. With a single optical flow measurement you can only detect an obstacle on one side, for example if you are approaching something from an angle. More optical flow measurements will allow more possibilities. for example, suppose you are computing two optical flows, one diagonally forward to the left and one diagonally forward to the right. You can do this with two sensors, or with just one sensor programmed to compute the two optical flows. You can then compute the divergence of the optical flow, or the difference between the two optical flows. This divergence will increase as you get close to an obstacle.

For more info, look here: http://centeye.com/technology/optical-flow/

And attached is a paper (there is an error in the paper though- equation 1 should have a sin function not cos).

Geof