Byung-Uk Park
HOUGH TRANSFORM FOR INTERIOR ORIENTATION IN DIGITAL
PHOTOGRAMMETRY
Sohn, Hong-Gyoo, Yun, Kong-Hyun
Yonsei University, Korea
Department of Civil Engineering
sohnl Q yonsei.ac.kr
ykh1207 Q yonsei.ac.kr
Yu, Kiyun
Deputy Director Team for National GIS
Ministry of Construction and Transportation, Korea
kiyun Q motc.go.kr
Jeong, Soo
Yonsei University Engineering Research Center, Korea
jeong@ yonsei.ac.kr
Working Group III/3
KEY WORDS: Hough Transform, Local Dynamic Thresholding, Digital Photogrammetry, Interior Orientation
ABSTRACT
It is known that Hough transform is insensitive to noise and the desired results can be acquired even in the damaged
image. Because of its applicability to the image with variations such as rotations or scale changes, Hough transform has
been widely used to detect the outlines of arbitrary-oriented and arbitrary-scaled objects such as straight lines or circles,
In this paper, we utilized Hough transform combined with the local dynamic thresholding method to accurately locate
the center location of fiducial marks. Using detected fidual marks and calibration data, it was possible to determine the
calibration coefficients to do interior orientation.
1 INTRODUCTION
Automatically locating the outline of arbitrary objects from digital image has been studied in digital
photogrammetry, pattern recognition and other disciplines. Extracting linear features or boundary of the objects is
primary task to perform higher level analysis of the input image. Hough transform has been widely used to identify
image features such as straight or curved lines (Ballard, 1981). It has also been applied to the variety of pattem
recognition of arbitrary oriented and shaped objects (Tzvi and Sandler, 1990; Xu et al.1990).
Many parameters and steps are included to calculate the 3-D coordinates of the object from two stereo image pairs.
As the first step of 3-D formation of the object, we perform the interior orientation to accurately establish the
relationship between pixel image coordinate system and image coordinate system. There are many interior orientation
parameters to be considered. They include atmosphere refraction, camera imperfections, film shrinkage, and scanner
error. Important element of interior orientation is to identify the central location of fiducial marks from input digital
image. In this paper, we utilized Hough transform combined with local dynamic thresholding method to automatically
determine the central location of fiducial marks. Using the detected fiducial marks and calibration data of fiducial marks,
the parameters from pixel coordinate system to photo coordinate system using affine transformation are determined.
2 ALGORITHM OF HOUGH TRANSFORM
Hough transform was first proposed by Hough (Hough, 1962). An example to detect straight lines using this method
was presented by Duda and Hart (1972). When many distributed points that seem to be in the straight line, the original
line can be traced along these points via Hough transform.
One pixel in image space is represented as one straight line in parameter space. Many pixels that are supposed to be
straight line in parameter space are mapped many lines in image space. Crossing point of these lines in parameter space
is detected as one straight line segment in the image space.
692 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
in the
Duda :
À strai
space |
paranx
model
Hough