Dumb experiments for road segmentation using Point Cloud from LiDAR

No high-math, neural networks or ready solution. Only dumb approaches like

• mark everything with specific height as road
• mark every occupacity grid cell with low height dispersion as a road
• start breadth-first search from the center (LiDAR location) and mark neighbour cells as road if there is a small height change etc
• Maybe, RANSAC

I'm not sure, but current (master) implementation is a dumb RANSAC