You are on page 1of 2

Master Resume

Position: [0.1300 0.1100 0.7750 0.8150]


Units: ’normalized’

Show all properties

make axes

Figura 1: Two views of a planar grid of points. a Frontal view, b oblique view

giving the virtual camera view shown in Fig. 1. The camera pose

>> Tcam = SE3(-1,0,0.5)*SE3.Ry(0.9)


Tcam =
0.6216 0 0.7833 -1
0 1 0 0
-0.7833 0 0.6216 0.5
0 0 0 1

>> cam.plot(P,’pose’,Tcam)
creating new figure for camera
h =
Axes with properties:
XLim: [0 1]
YLim: [0 1]
XScale: ’linear’
YScale: ’linear’
GridLineStyle: ’-’
Position: [0.1300 0.1100 0.7750 0.8150]
Units: ’normalized’
Show all properties
make axes

results in an oblique view of the plane as shown in Fig. 2. We can clearly see the effect of perspective
projection which has distorted the shape of the square – the top and bottom edges, which are parallel lines,
have been projected to lines that converge at a vanishing point.
The vanishing point for a line can be determined from the projection of its ideal line. The top and
bottom lines of the grid are parallel to the world x-axis or the vector (1, 0, 0). The corresponding ideal
line has homogeneous coordinates (1, 0, 0, 0) and exists at infi nity due to the fi nal zero element. The
vanishing point is the projection of this vector which is (1 803, 512) and just to the right of the visible
image plane.

PUC-Rio 5 ELE
Master Resume

>> cam = CentralCamera(’focal’,0.015,’pixel’,10e-6,’resolution’,[1280 1024],’centre’, [640 512], ’n


cam =
name: mycamera [central-perspective]
focal length: 0.015
pixel size: (1e-05, 1e-05)
principal pt: (640, 512)
number pixels: 1280 x 1024
pose: t = (0, 0, 0), RPY/yxz = (0, 0, 0) deg

which displays the parameters of the camera model including the camera pose T. The corresponding
non-homogeneous image-plane coordinates of the previously defined world point are

>> cam.project(P)
ans =
790
712

3. Camera Matrix
Combining Eq 4 and Eq. 5 we can write the camera projection in general form as

 1  
ρw 0 u0 f 0 0 0
p̃ =  0 1
ρh v0   0 f 0 0 (0 TC )−1 P̃ (7)
| {z }
0 0 1 0 0 1 0 extrinsic
| {z }
intrinsic
= KP0 0 T−1
C P̃ (8)
= C P̃ (9)

where all the terms are rolled up into the camera matrix C. The terms ρfW and ρfh h are the focal length
expressed in units of pixels. This is a 3×4 homogeneous transformation which performs scaling, translation
and perspective projection. It is often also referred to as the projection matrix or the camera calibration
matrix.
The projection can also be written in functional form as

p = P(P, K, ξC ) (10)

where P is the point coordinate vector in the world frame. K is the camera parameter matrix and
comprises the intrinsic parameters which are the innate characteristics of the camera and sensor such as
f, ρW , ρh , u0 and v0 .
ξC is the pose of the camera and comprises a minimim of six parameters-the extrinsic parameters - that
describe camera translation and orientation in SE(3).
There are 5 intrinsic and 6 extrinsic parameters – a total of 11 independent parameters to describe
a camera. The camera matrix has 12 elements so one degree of freedom, the overall scale factor, is un-
constrained and can be arbitrarily chosen. In practice the camera parameters are not known and must be
estimated using a camera calibration.
The camera intrinsic parameter matrix K for this camera is
>> cam.K
ans =
1.0e+03 *

1.5000 0 0.6400
0 1.5000 0.5120
0 0 0.0010

PUC-Rio 3 ELE

You might also like