RESULTS OF THE GERMAN COMPARISON TEST FOR DIGITAL POINT OPERATORS 
Thomas Luhmann 
Fachhochschule Oldenburg 
Laboratory for Photogrammetry 
Germany 
Intercommission Working Group V/IIl 
KEY WORDS:  Close-Range Photogrammetry, Digital Point Measurement, Sub-Pixel Accuracy, Simulation 
ABSTRACT: 
A test procedure has been developed in order to compare different digital point measurement techniques. For this test a 
number of synthetic and real images with various circular point patterns have been generated and submitted to 20 
different institutions throughout Germany, Switzerland and Austria (universities and system developers). One goal of 
this test is to compare the suitability of different point operators under equal conditions. A second goal is to verify the 
current limits of sub-pixel point positioning which, by numerous authors, have been reported down to a few thousands of 
a pixel. The investigation of the image test material is based on a 2-D (image space) and 3-D (object space) analysis. 
The comparison study reports of the results of the following point operators: structural edge-based ellipse operators, 
center-of-gravity operators, template matching operators and least-squares matching operators. 
Using synthetic images the results show that a sub-pixel accuracy of 1/100 of a pixel size can be achieved by edge- 
based ellipse operators. With real images taken from a fully-digital high-resolution CCD camera the mean accuracy in 
image space is limited to 2/100 of a pixel (under laboratory conditions). These results are obtained from test field 
calibration using self-calibrating bundle adjustment. They correspond to a relative accuracy of about 1:140.000 with 
respect to the largest object dimension. 
KURZFASSUNG: 
Dieser Bericht fafit die wesentlichen Ergebnisse einer Vergleichsstudie zusammen, die die Leistungsfähigkeit 
verschiedener digitaler Punktmeßalgorithmen untersucht. Dazu wurden diverse synthetische und reale Bilder erzeugt, in 
denen verschiedene kreisförmige Punktmuster abgebildet sind. Dieses Testmaterial wurde an 20 wissenschaftliche und 
kommerzielle Institutionen in Deutschland, Österreich und der Schweiz verteilt, die die abgebildeten Punkte mit ihren 
eigenen Verfahren gemessen haben. Zum einen sollte in der Studie ein Vergleich verschiedener PunktmeRverfahren 
unter identischen Bedingungen durchgeführt werden. Zum anderen sollte die Grenze der Subpixel-Mefigenauigkeit 
evaluiert werden, die in verschiedenen Publikationen bereits auf wenige tausendstel Pixel definiert wurde. Der 
Vergleichstest stützt seine Ergebnisse sowohl auf eine 2D- (Bildraum) als auch eine 3D- (Objektraum) Analyse. 
Folgende Algorithmen konnten untersucht werden: strukturierte kantenbasierte Ellipsenoperatoren, Schwerpunkt- 
Operatoren, Template-Matching-Verfahren und Kleinste-Quadrate-Verfahren. 
Mit synthetischen Bildern werden Genauigkeiten von bis zu 1/100 Pixel erreicht (Ellipsenoperator). In den mit einer 
volldigitalen, hochauflósenden CCD-Kamera aufgenommenen realen Bildern werden Bildmef&genauigkeiten von 2/100 
Pixeln erreicht (Laborbedingungen). Diese Ergebnisse wurden im Rahmen einer Testfeldkalibrierung erreicht. Sie 
entsprechen einer Relativgenauigkeit von ca. 1:140.000, wenn die grófàte Objektausdehnung zugrunde gelegt wird. 
1 OBJECTIVES accuracy acceptance test procedure. In the world of CMM 
applications these tests have been developed and they 
In recent years great progess in development and are used permanently. In industrial photogrammetry there 
application of digital close-range systems can be is not only a lack of such a common acceptance test, or 
recognized. These systems are mostly used in industrial but also a standard of common language in terms of 
application, especially in the fields of quality assurance, ^ accuracy and system performance. 
production process control and research and 
development. Most systems are using high-resolution In this context one objective is the investigation of sub- 
digital cameras, artificial point targeting, precise digital pixel accuracy of digital point measurement methods. 
image measuring algorithms and bundle adjustment Image acquisition and processing are fundamental steps 
techniques. Recent publications show that digital in digital close-range photogrammetry that basically affect 
photogrammetric systems are meanwhile accepted in the resulting system accuracy. Object information lost in 
industry where they compete with classical measuring image space cannot be reconstructed afterwards. 
devices such as coordinate measurement machines 
(CMM) (Luhmann 1995). In 1995 the author initiated a comparison test for the 
evaluation of different digital point measurement 
In this environment it is essential not only to achieve high techniques. For this test a number of synthetic and real 
accuracies but, to have a proven kind of system and images with various point patterns were generated and 
324 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996 
  
submittec 
to comp: 
under ec 
current | 
numerou: 
thousand 
Up to no 
the follo 
ellipse © 
matching 
operators 
the test 
internatic 
and resu 
The inve: 
2-D (ima 
D analy: 
centers | 
values gi 
point pat 
The 3-D 
images | 
field wa 
targets v 
analysis 
coordina 
coordina 
calibratic 
The test 
given ob 
accuracy 
done at 
Approx. 
throughc 
the test 
participa 
Is