Sie sind auf Seite 1von 31

Sensitivity Analysis

of Genetic Algorithm Based


Camera Calibration
Project done by
A. Abinesh (05G01)
M. Arul Ram (05G03)
B.E. Final Year

Guided by
Mr. C. Muruganantham,
Associate Professor,
Mechanical Engineering Department
ABSTRACT

In this project, a robust camera calibration based on the
linear camera model is performed. In this approach
Genetic algorithm is used as a tool for performing
camera calibration. Also the distorted image plane error
is calculated as a performance measure, it provides
accuracy for camera calibration.
Sensitivity analysis of the parameters is performed in
which a known percentage of error is given to the
parameters, to find which parameter has greater
influence on the camera calibration.
The results indicate that the principal point is highly
sensitive in machine vision applications and all other
parameters are relatively less sensitive.
CAMERA CALIBRATION

Camera Calibration is the process of
determining the internal geometric and optical
characteristics (Intrinsic parameters) and/or the
3-D position and orientation of camera relative to
a chosen world coordinate system (Extrinsic
parameters)

INTRODUCTION
EXTRINSIC PARAMETERS
TRANSLATION ALONG X
AXIS
TRANSLATION ALONG Y
AXIS
TRANSLATION ALONG Z
AXIS
ROTATION ALONG X AXIS
ROTATION ALONG Y AXIS
ROTATION ALONG Z AXIS

INTRINSIC PARAMETERS
FOCAL LEGTH OF THE LENS
SCALING FACTOR ALONG X AXIS
SCALING FACTOR ALONG Y AXIS
CENTRE OF IMAGE PLANE(U0,V0)
Coordinate Systems
World Coordinate System: Its a known
reference coordinate system with respect to
which we calibrate the camera.
Camera Coordinate System: Its a coordinate
system with its origin at the optical center of the
camera.

Geometry of image formation
MATHEMATICAL FORMULATION
Projection from world to pixel reference frame using all
camera parameters without considering distortion is given
by:

u 1/sx 0 u
o
f 0 0 r
1
r
2
r
3
t
X
X
v = 0 1/sy v
o
0 f 0 r
4
r
5
r
6
t
Y
Y
1 0 0 1 0 0 1 r
7
r
8
r
9
t
Z
Z
0 0 0 1 1


Rotation &
Translation
Projection
Scaling &
Translation
World
point
Image point
Need for calibration
The basic need is to find an equation linking the
camera reference frame with the image
reference frame (Link I), and another equation
linking the world reference frame with the
camera reference frame (Link II).
Solving this system, results in linking the link
between world reference frame and image
reference frame.
Finding Link I and Link II are equivalent to
finding the camera's characteristics, also known
in computer vision as the camera's extrinsic and
intrinsic parameters.
Camera calibration
Mainly, two categories:
1. Photogrammetric calibration: use reference
objects with known geometry
2. Self calibration: only assume static scene, e.g.
structure from motion

We prefer photogrammetric technique because
calibration can be done with very good precision
and efficiently
DEFINITION OF THE GENETIC
ALGORITHM
The genetic algorithm is a probabilistic search
algorithm that iteratively transforms a set of
mathematical objects each with an
associated fitness value, into a new
population of offspring objects using the
Darwinian principle of natural selection and
using operations that are patterned after
naturally occurring genetic operations, such
as crossover and mutation.
Selection is done on the basis of relative fitness
and it probabilistically culls solutions from the
population that have relatively low fitness.
Crossover is a structured yet stochastic operator
that allows information exchange between
candidate solutions.
Mutation insures against the permanent loss of
genetic material during the selection process.

WHY TO USE GA FOR CAMERA
CALIBRATION

A key issue that arises in this approach is the
extremely large search space caused by the
presence of uncertainties.
This presents a serious challenge to
conventional minimization procedures since
there may be several local minima, and the
choice of starting point will determine which
minimum the procedure converges to, or
whether it will converge at all.
If starting points are far away from the desired
minimum, the traditional optimization techniques
could converge erroneously
SPECIFICATIONS
CCD Camera
Scanning System Progressive Scan
Pixel Clock 20.25 MHz
CCD Sensor Monochrome 2/3 IT CCD
Sensing Area 8.7 mm X 6.9 mm
Picture Elements 1300 X 1030
Cell Size 0.0067 mm X 0.0067 mm
Resolution 1040 X 1030 TV Lines
Result For Calibration
Results for various population
PARAMETER POP=25 POP=50 POP=75 POP=100 POP=125 POP=1000
z-axis -1.112 -1.112 -1.112 -1.112 -1.112 -0.854
x-axsis -0.339 -0.339 -0.339 -0.339 -0.339 -2.336
y-axis 1.997 2.518 -2.518 -2.518 -2.518 -2.405
focal length 24.942 24.354 24.354 24.354 24.354 24.354
scale factor, x 1.19 1.19 1.19 1.19 1.19 0.376
scale factor, y 0.678 0.678 0.678 0.678 0.678 1.352
translation, x 0.9 -0.9 -0.836 -0.836 -0.836 -0.836
translation, y -0.846 0.652 -0.634 -0.634 -0.634 -0.634
translation, z 50.88 50.88 50.88 50.88 50.88 36.48
uo 649.86 649.86 641.4 641.4 641.4 645.62
vo 518.063 518.063 508.547 508.547 508.547 519.844
DISTORTED IMAGE PLANE ERROR:

One of the first questions about camera
model is how accurately it captures the
cameras imaging behavior.
This information is necessary both for
measuring progress during model
calibration and estimating the performance
or accuracy of any application the model is
used in.
Distorted image plane error gives how
accurately camera model captures the cameras
imaging behavior.
Given the measured coordinates of a point in the
object space (xw, yw, zw ) and the measured
position of the points image in the frame
grabber (Xf, Yf) we can define an error metric for
the model anywhere along the models chain of
coordinate transformations.
One obvious error metric is the difference
between the position of a points image we
measure and the position the camera
predicts. Thus it can be defined the
distorted image plane error (DIPE) as

DIPE=

2 2
( ) ( ) f f f f X X Y Y


POPULATION SIZE DIPE
25 43.667
50 43.546
75 31.949
100 31.949
125 31.949
1000 39.903
SENSITIVITY ANALYSIS:
In LP, the parameter of the model can
change within certain limits without
causing the optimum solution to change.
This is referred to as sensitivity analysis
In LP models, the parameters are usually
not exact. With sensitivity analysis, we can
ascertain the impact of the uncertainty on
the quality of the optimum solution.
Sensitivity analysis is used to determine
how sensitive a model is to changes in
the value of the parameters of the model
and to changes in the structure of the
model. By showing how the model
behavior responds to changes in
parameter values, sensitivity analysis is a
useful tool in model building as well as in
model evaluation.

For example, for an estimated unit profit of
a product, if sensitivity analysis reveals
that the optimum remains the same for a
10% change in the unit profit, we can
conclude that the solution is more robust
than in the case where the difference
range is only 1%.

Sensitivity analysis helps to build
confidence in the model by studying the
uncertainties that are often associated with
parameters in models
Sensitivity analysis allows one to
determine what level of accuracy is
necessary for a parameter to make the
model sufficiently useful and valid.
SENSITIVITY ANALYSIS
31.4
31.5
31.6
31.7
31.8
31.9
32
32.1
32.2
32.3
-10 -8 -6 -4 -2 0 2 4 6 8 10
error
D
I
P
E
sx
tx
ty
tz
za
xa
ya
focal
SENSITIVITY ANALYSIS
0
10
20
30
40
50
60
70
80
90
100
-10 -8 -6 -4 -2 0 2 4 6 8 10
ERROR
D
I
P
E
cx
cy
CONCLUSION

Using genetic algorithm JAI-CV M1 Progressive
Scan camera is calibrated and the camera
parameters are estimated.
At the completion of sensitivity analysis it is found
that except the principal point (u0, v0), all the
parameters are less sensitive as their variation is
minimum of 1% of the DIPE.
Thus it is concluded that while calibrating a
camera or in the application of camera calibration
utmost care should be taken in finding the
principal point,
REFERENCE

Fabio Remondino, Niclas Boerlin, Photogrammetric Calibration Of
Image Sequences Acquired With A Rotating Camera, International
Archives Of Photogrammetry Remote Sensing And Spatial Information
Science, Vol. 34, Part 5, 19-22 Feb 2004
Motilal Agrawal and Larry S. Davis ,Camera calibration using
spheres: A semi-definite programming approach, Ninth IEEE
International Conference on computer vision, Vol. 2, 2003, pp-782
Qiang Ji &Yongmian Zhang, Camera calibration with genetic
algorithms, IEEE Transactions on Systems, Man and Cybernetics,
Part A, Vol. 31 Issue: 2, (March 2001), pp. 120 130.
Roger Y. Tsai, An efficient and accurate camera calibration
technique for 3D machine vision, Proceedings of IEEE Conference on
Computer Vision and Pattern Recognition, Miami Beach,
February,1986, pp.364-374.
Savii.G.G,Camera Calibration Using Compound Genetic-
Simplex,Journal of Optoelectronics and Advanced Materials Vol.
6,No. 4,December 2004,p. 1255 1261.
Zhang, Z.,A flexible new technique for camera calibration,IEEE
Transactions on Pattern Analysis and Machine Intelligence, Vol. 22,
no. 11, (Nov. 2000), pp. 1330 1334.

Xiaoqiao Q. Meng and Z Y. Hu, A new easy camera calibration
technique based on circular points, (2003). Pattern Recognition.
Vol. 36, No. 5, pp. 1155-1164.
Kalyanmoy Deb Optimization for Engineering Design Algorithms
and Examples,. Prentice-Hall of India Private Limited, ISBN-81-
203-0943-X. pp 290-319.
Reg G. Wilson Modeling and Calibration of Automated Zoom
Lenses, proceeding of the SPIE #2350: videometricsIII, October,
1944, PP.170-186.
Hamdy A. Taha, Operations Research An Introduction,
Prentice-Hall of India Private Limited. ISBN-978-81-203-3034-6.
pp 123-130.
Hati S. & Sengupta S., Robust Camera Parameter Estimation
using Genetic Algorithm, Proceedings of the IEEE International
Conference on Systems, Man and Cybernetics, vol.4, (12-15 oct,
1999), pp.943-947.

Das könnte Ihnen auch gefallen