Vision 1: Camera models + Calibration

Question 1

Name some static and perspective monocular depth queues

Answer 1

1. Static (pictorial) cues
   
   1. Occlusion
   2. Relative size
   3. Aerial (Atmospheric)
   4. Shading
2. Perspective
   
   1. Linear perspective
   2. Texture gradient
   3. Relative Height
   4. Cast shadows/shading

Question 2

How do you calculate the depth of a point using two cameras facing the same direction?

Answer 2

In order to find the depth of a point with two cameras facing the same direction, you take the length between the optical centers, divide it by the difference in x positions in the images, and multiply it with the focal length of the camera.

Question 3

in a matrix:
[alpha_u, s, u_0;
0 , alpha_v, v_0 ;
0,0,1]
What are the variables alpha_u, alpha_v,s,u_0,v_0

Answer 3

alpha u and v are the pixel reselution of the image plane, s the skew, u_0 and v_0 are the optical centre.
The skew is often zero.

Question 4

What are some diffrent apporches to calibration?

Answer 4

Using a well known 3D structure at known 3D position.
Auto-calibration(Needs multiple veiws, cna be achived by moving the camera)
Planar object at multiple locations, for example multiple images of a checkerbord

Question 5

What size is P? (P = KAH)

Answer 5

3x4 matrix

Question 6

If we have M (a 3D Point) and P, can we compute m (2D point on image plane)? If we can, how is it done?

Answer 6

m = P * M

Question 7

If we have m (2D point on image plane) and P, can we compute M (3D point)?

Answer 7

No, but we can find a the vector the point lies on