I am trying to detect a crosswalk pattern on the road by using Hough transform, and then selecting lines that separate between black and white regions in a binary image. The stripes are detected fine, but the problem is that a lot of other unwanted objects provide false positives, as you can see in the image:
*You can ignore the non-parallel lines, they are not taken into account.
I wish to ignore things such as the wall and the taxi that are also 'parallel' to the crosswalk stripes.
So my question is: Is there an algorithmic way to detect the ground plane(road in this case), even partially, in order to improve detection?