computer vision - Essential Matrix from Fundamental Matrix in OpenCV -


i've computed fundamental matrix of stereo pair through corresponding points, found using surf. according hartley , zisserman, essential matrix computed doing:

e = k.t() * f * k 

how k? there way compute e?

i don't know got formulae, correct 1 e = k'^t . f . k (see hartley & zisserman, §9.6, page 257 of second edition)

k intrinsic camera parameters, holding scale factors , positions of center of image, expressed in pixel units.

    | \alpha_u     0     u_0 | k = |    0      \alpha_u v_0 |     |    0         0      1  | 

(sorry, latex not supported on so)

edit : values, can either:

  • calibrate camera
  • compute approximate value if have manufacturer data. if lens correctly centered on sensor, u_0 , v_0 half of, respectively, width , height of image resolution. , alpha = k.f f: focal length (m.), , k pixel scale factor: if have pixel of, say, 6 um, k=1/6um. example, if lens 8mm , pixel size 8um, alpha=1000

computing e

sure, there several of ways compute e, example, if have strong-calibrated rig of cameras, can extract r , t (rotation matrix , translation vector) between 2 cameras, , e defined product of skew-symmetric matrix t , matrix r.

but if have book, of inside.

edit noticed, there wikipedia page on topic!


Comments

Popular posts from this blog

php - render data via PDO::FETCH_FUNC vs loop -

c++ - OpenCV Error: Assertion failed <scn == 3 ::scn == 4> in unknown function, -

The canvas has been tainted by cross-origin data in chrome only -