detect road shape on image without lanes lines

asked 2017-03-24 22:53:31 -0500

WiVM gravatar image


I am trying to find a way to detect the shape of the road surface. I have experimented a lot with the examples on lane detection, but the issue is that the contrast of the concrete surface and road sides (often grass) is not enough especially when there are no side lines as you have in lanes. A good example is the image below. It uses background abstraction to find the car, but I actually need to road surface shape (preferably without the lighter gray side ways) also without having painted lane lines like your normally have.

Many thanks

edit retag flag offensive close merge delete


Hi, I am still trying to find an approach to solve this issue. Base on a document I found about road edge detection in infrared images, it was suggested to down sample the image using the grassian pyramid technique and detect the convergence common point on the horizon with hough lines. As expected there will be an adaptive threshold as the vehicale is moving. Is someone having an example of these techniques? Or maybe is aware of a project that does exactly this? It appears to be a technique that is used in the automotive industry to detect pedestrians as they are suppose to be on the road sides. My additional issue is that I am a VB .NET developper and the conversion from C++ is not an easy one. Thank you!

WiVM gravatar imageWiVM ( 2017-03-30 02:11:46 -0500 )edit