You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Proceedings, XXth congress

Mehdi Ravanbakhsh, Saeid Sadeghian
Research Institute of National Cartographic Center (NCC), Tehran, Iran, P.O.Box:13185-1684
(Ravanb, Sadeghian)@ncc.neda.net.ir
Commission II/IV
KEY WORDS: Orientation, Accuracy, digital, Transformation, Space, Algorithms
This paper concerned with a comparative study and implementation of Interior orientation manually and automatically for KFA-
1000. In manually method, firstly, we reproduced the image to its reduced size photographically from the original image to be
measurable in a mono comparator from 30*30 cm to 23*23 cm which proceeds by measuring of photo coordinates of four pieces of
KFA-1000 photo. After computation, it was realized that geometric distortion due to photography is high so the chosen method is to
make a photographic reproduction of the image in suitable pieces, measuring with traditional instruments and pining the pieces
together before calculation. In the second method, we scanned the space photo and used digital image processing techniques in
which CCF (Cross correlation Function) and BinaryCCF were used to approximate positioning followed by quadratic surface fitting
for the purpose of precise fiducial centre determination. In both methods, we applied three types of transformation called Conformal,
Affine and 2D projective to the algorithms.
The final results showed that we could achieve sub-pixel accuracy of nearly 10 micrometer for 25 micrometer pixel size image by
using 2D projective transformation which is approximately equal to the accuracy achieved in manual method. With respect to
differences between aerial and KFA-1000 photos, we considered those differences and applied them to the algorithms used in both
fields (aerial and space) particularly different background at location of each fiducial mark in KFA-1000 which made us to resample
a suitable template while in aerial photos backgrounds in any photo are similar with good contrast which guarantee the robustness of
the algorithms. We also examined the effect of pre-processing techniques on the robustness of algorithm and took into account the
1. INTRODUCTION defined by fiducial marks positions. If the relationship between
scanner and camera coordinate system remain constant, inner
Image orientation is a prerequisite for any project including 3D orientation will be eliminated from digital photogrammetric
computation as it is a complicated and time-consuming process stages. We have not had the chance of using such digital
which making it automatic helps us to open a broad range of ^ cameras to prepare high resolution images yet.
applications. Due to its decisive importance, image orientation We would be able to design an algorithm to achieve the goal of
has always been a focus of attention in photogrammetry. Digital performing interior orientation automatically if it is possible to
photogrammetry holds the promise of completely automating ^ have a good knowledge of location, shape, illumination
the process of image orientation using image processing and distribution and sizes of fiducial marks in digital scanned KFA-
image analysis techniques. It is interesting to note in this 1000 photos. With respect to fully automation procedure
context that in the computer vision terminology "image ^ concept, the algorithm should be able to include features listed
calibration" sometimes already includes image orientation. below:
Image orientation refers to the determination of parameters |-using images with different pixel size
describing specific photogrammetric models for mapping 2-using positive or negative images
geometric primitives such as points, lines, and areas from one 3-using mirror images
coordinate system to another one. Thus, image orientation 4-using rotated images
belongs to the class of coordinate transformation problems. 5-using images scanned in different scanners
Among orientations (inner, relative and absolute) inner
orientation has a special importance as any inaccuracy will 3. DESIGNING AND IMPLEMENTING
affect next stages in photogrammetric processes.
With respect to the kind of input data, several methods can be
taken into account, however, two distinct algorithms performing
two major successive stages of localization and precise
measurement differently and other minor six stages commonly.
2. INNER ORIENTATION AND AUTOMATION The six common stages are:
FEATURES l-extracting image patches
2-resampling the template
In inner orientation process, we establish a geometric 3- Image pyramid derivation
relationship between photo coordinate system and instrument 4- Detection of the orientation of the image
coordinate system. In metric imaging system, photo coordinate 5-positive-negative recognition
system is defined by fidcual marks but in field of digital 6-Estimation of the transformation parameters
photogrammetry instrument coordinate system is replaced by ^ The whole procedure of automatic interior orientation (AIO) is
pixel coordinate system. Image coordinate system is defined by shown in Fig.1.
the matrix including grey values but photo coordinate system is