On the paper:

• Grayscale: Did you try splitting the colors and seeing if just using the

grayscale versions of G, R, B helped any better?

• Blurring: Did you try anything other than 5x5

• Hough Transform: Why probabilistic and not just the Hough transform

How about adjusting the parameters for this transform and viewing

results?

• While you can adjust and tune for a given set of images, how would you

handle different conditions, lighting for example?

• I'd have picked a different color than green to highlight the lines with the

grass as it blends in with the grass.

• How about examples with indoor conditions? How did things fair with

those?

• The overall goal is to produce an algorithm that runs on the pi. Any

suggestions about how you might modify things to run on it instead of

your laptop?