MULTLIMAGE BASED CAMERA CALIBRATION WITHOUT CONTROL POINTS
XIE Wenhan,ZHANG Zuxun,ZHANG Jianging
School Of Remote Sensing Information Engineering, Wuhan University, Wuhan, China, 430079
xiewenhan(@263 .net
PS WG V/1
KEY WORDS: Camera calibration, Close range photogrammetry, Threcdimensional reconstruction, Architecture photogrammetry,
Bundle adjustment
ABSTRACT:
In recent years, many researches focus on building 3D object model in the fields of computer vision and photogrammetry,
S
per
and camera calibration accordingly become the key problem. In this paper, the characteristic and shortage of the theory
about calibration with vanishing points are exploited. The error was analyzed which is resulted from calibration based on
vanishing points. Furthermore, a novel approach for camera calibration utilizing vanishing points in multi images is
proposed. This approach overcame the demerit of the theory mentioned above; therefore the resulting precision of interior
and exterior orientation parameters was improved.
1. INTRODUCTION
The research regarding obtaining 3D information of objects in
nature from 2D images is always a hotspot in the fields of
computer vision and photogrammetry, and the key problem is
whether interior orientation parameters, which determine the
geometry relationship between 2D images and 3D objects, can be
obtained accurately. The process of obtaining interior
orientation parameters of camera is in terms of camera
calibration. For camera calibration the simplest imaging
geometry model—linear imaging model of camera is commonly
adopted. It is essential that the camera was calibrated before 3D
object reconstruction.
In the field of photogrammetry, there are many conventional
methods for camera calibrating such as Optical Laboratory
Calibration, Test Range Calibration, On the Job Calibration and
Stellar Calibration etc. the characteristic of these methods is high
precision and demanding of the surveying job with high precision.
However, most of them need place a number of reference objects
in the front of camera, it makes the operation procedure not
flexible, and in many special environments, there are no
reference objects or test range can be used. With the rapid
popularization and development of digital camera, many
researchers try to solve this problem with a novel calibrating
method without reference objects. For some applications the
accuracy of camera calibration need not to be very high, and
sometimes the focal length of camera is even various with
shooting, thus it can't be calibrated beforehand. Therefore, the
camera calibrating method without reference
objects—calibration based on vanishing points is applied
broadly.
2. THEORY AND ADJUSTMENT MODEL UTILIZING
VANISHING POINTS OF SINGLE IMAGE
Under the perspective projection, lines, which are parallel in
space, will converge to a point in the image plane, this point
called vanishing point. Therefore, the surface image of objects
taken by camera provide rich geometric line information (such as
the parallel and perpendicular outlines of the building surface),
the camera calibration can be fulfilled by fully utilizing vanishing
points.
2.1 Single-view calibration with vanishing points
Many studies about camera calibration are based on vanishing
points. Prof. Heuvel in Delft University was engaged in the study
of Architectural Photogrammetry, where he calibrated camcra
using vanishing points [5] He mainly calibrated interior
orientation parameters of camera utilizing the image with
three-point perspective, accordingly linc photogrametry was used
to solve the problem of 3D model reconstruction. Roberto
Cipolla in University of Cambridge used strict geometrically
intuitive condition of object (parallelism and orthogonality) to
calibrate the intrinsic and extrinsic orientation parameters of the
cameras and to recover Euclidean models of the scene from only
two object images taking from arbitrary positions [6].
Schuster et al investigated the usefulness of vanishing points for
steering(navigating) a robot, the objective is to steer a mobile
robot based on the vanishing points of parallel lines in its
environment [7]. Tsai employed a hexagon as the calibration
target to generate a vanishing line of the ground plane from its
projected image, and calibrated parameters include the
orientation, the position, and the focal length of a camera [9].
Internati
Prof. Nev
camera b
vanishing
motion |
proposed
informati
only with
Figi
Suppose
image rn
perspecti
Fig.l, Th
formed b:
f ai
Three ext
tan @ =
tan o -
tan K =
2.2 Adj
Vanishing
paramete:
points. M
Barnard f
Gaussian
E. Luttor
[12]; Can
points fi
intersectis
