TOWARDS A HIGHER LEVEL OF AUTOMATION
FOR SofiPlotter™
Yan Lue, Senior Photogrammetric Engineer
Vision International
A division of Autometric, Inc.
502 Earth City Expressway, Suite 203
Earth City, MO 63045, USA
lue(@stl.autometric.com
Commission III, Theory and Algorithms
KEY WORDS: Photogrammetry, Automation, Matching, Interior Orientation, Feature extraction, Tie Points selection
ABSTRACT:
In this paper we present the efforts on two important aspects of automation for the digital photogrammetry system: Automatic
Interior Orientation (470) and Automatic Tie Point Selection (A47PS). The latter one covers the traditional point selection, transfer
and mensuration. Both modules are available in the SoftPlotter™ softcopy photogrammetric workstation from Vision International.
A fully automatic .4/0 module based on template matching, together with LSM techniques suitable for most commercially available
aerial cameras, has being running on many SoftPlotter™ for over a year. A success rate of one hundred percent for fiducial
recognition with very high accuracy was achieved on several hundred test frames. Thousands of frames have been processed with
AIO by SoftPlotter users with satisfactory results. The processing time averages only a few seconds per frame with eight fiducials.
Another fully automatic module A7PS is based on a series of sophisticated image matching algorithms and strategies. The beta
version has been applied to some practical projects. The A7PS results for a variety of strips covering image scale from 1:3400 to
45000 with all kinds of different textures, including urban area with strong distortion, forest, mountain, wetland, desert, rivers,
lakes etc., are very promising. The feedback from the real world is very positive too. The averaging success rate was about 85
percent. The processing time was less than a half minute per frame with points on nine standard positions.
0. INTRODUCTION
Digital photogrammetry (DP) is coming out from its research
and development period and into day-to-day practice.
Commercial photogrammetric companies are rapidly turning to
digital photogrammetric workstations and systems to meet
production requirements. Traditionally, the most time-
consuming and labor-intensive task is the mensuration of the
all kinds of points, such as fiducial marks, ground control
points, tie points. Somewhat surprisingly, however, while
most commercially available digital photogrammetric systems
have provided automatic photogrammetric data collection and
product generation for DEM and orthophoto for quite a while,
the most fundamental operations of photogrammetric
processing like interior orientation, tie points selection and
mensuration needed by aerial triangulation (AT) are still often
performed manually as on an analogue or analytical plotter.
In this paper we present our efforts on two important aspects
of automation for the digital photogrammetry system:
automatic interior orientation (4/0) and tie point selection
(ATPS). The first part gives a brief description on the basic
concepts and technical strategies and then introduces some
results on thousands of aerial photos taken by eight popular
aerial cameras (Wild RC8, RC10, RC20, RC30, Jena LMK
1000, 2000, Zeiss RMK TOP 15, RMK A 15/23, 30/23) (Lue,
1995). The second part describes methods and strategies for
ATPS procedure, which shares a similar philosophy and
methodology but uses more sophisticated strategies and
478
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
organization. A summary about the ATPS results on 16 data
sets with different image scales, different scanning resolutions
and all kinds of different textures, including urban area with
strong distortion, forest, mountain, wetland, desert, rivers,
lakes etc. is also provided.
Some technical and practical points is discussed in the
successive section, then a conclusion is drawn in the last part
of the paper.
1. FULLY OPERATIONAL AUTOMATIC
INTERIOR ORIENTATION (AIO)
In photogrammetric processing image measurements of all
points, no matter if they are tie, ground control, feature or
terrain elevation points must be reduced to the camera system
central perspective coordinates. For the DP processing a
hardcopy film is arbitrarily put on a scanner for digitizing. A
set of transformation parameters is therefore needed to derive
the film coordinates from the scanning ones or vice versa. The
tasks of recognizing and measuring fiducials for deriving the
transformation parameters between the scanned image and
camera coordinate systems comprise the two basic tasks of
interior orientation. As well known, the latter one has been.
well established. The automation of the interior orientation
needs to concentrate primarily only on the former
Interi
the fi
the r
Vario
SoftP
auton
inten
The a
matcl
ERD:
augm
accur
follox
litera
1.1 B
Basi
The
imag.
squai
Take
Com]
for tl
know
prior
and 1
In ge
and/(
diffei
shape
fiduc
exter
diffe:
scant
Figui
bette
imag
Clea
perfc
thea
a qu
pred
pyrai
use.
Tem
toge
It is
The
appr
bette
atter
relia
guar
For |
read
effic
pyra
smal
proc