Hasta La Motion!

I am doing 2D to 3D conversion in nuke. I need to extract a depth map of a live action shot. How can I do that? I tried the easy way in 2D making roto, ramp and fill my roto with different gray level. and I try with a camera to project my roto on card on different z positions in 3D space; The question is, when I look at my depth layer from the scanlinerender, how can I be sure my differents cards are at the good positions the one from the other. For example, I have a landscape, with mountains on the back, people on front; so I made layers for people and mountains perpendicular to the cam and one for the ground. How to I make sure my people stick to the ground and the contact of my people and my moutains with the ground card is accurate? I means where do I put those two layers in z according to the ground. I don’t know if I make myself clear.