Question

I am taking part in a line following robot competition where we use a camera in order to follow line.

My algorithm is:

mean filter
--> Gaussian filter
--> find centroid
--> kalman filter
--> do pid for steering
--> do pid for speed of back motor

and the robot performs 98% perfectly.

My issue is that after all this filtering, sometimes I see some huge error in the reading of the camera when the robot moves, which cause small oscillation of the robot.

How can I eliminate this error?
(this small error in high speed is a big issue)

Answer 1

Probably setting up the buffers before image processing can work.
Assume of three buffers (image pages saved continuously from camera)… page1, page2, page3.
Use page1 to process the image… Then get an average image by comparing both page2 and page3, and put this resulting image on page1… then repeated the image processing from page1 again… and so on.