CALIBRATION AND ACCURACY ASSESSMENT OF A 
MULTI-SENSOR ONLINE-PHOTOGRAMMETRIC SYSTEM 
Robert Godding 
Institute for Photogrammetry 
and Image Processing (IPB) 
Technical University Braunschweig 
Germany 
Thomas Luhmann 
Leica Heerbrugg PMU 
Research and Development 
Switzerland 
ISPRS Commission V 
ABSTRACT: 
The Programmable Optical Measuring System (POM) was developed for use in non-contact 3D measurement 
applications in industrial metrology. The system consists of different digital imaging sensors (high-resolution and 
standard CCD cameras), a digital rotary table and active light sources. The rotary table enables views from all around 
the object. 
This paper reports on the calibration of the complete system, which includes the determination of exterior orientation as 
well as interior calibration of cameras and rotary table. The complete set of orientation values is calculated by self- 
calibrating bundle adjustment. 
3D object information is obtained from spatial intersection or 3D element adjustment. The accuracy which can be 
achieved under practical conditions depends on the quality of calibration, the photogrammetric network configuration, 
and the quality of image measurement. Accuracy estimation is verified by independent measurements using calibrated 
reference objects such as scale bars. It is shown that an overall accuracy of 0.1mm for an object size of 2m can be 
obtained. 
Key Words: on-line photogrammetry, digital high-resolution camera, self-calibration, rotary table 
1. INTRODUCTION 
LEICA's Programmable Optical Measuring System 
(POM) is a flexible multi-sensor digital photogrammetric 
system for automated industrial applications. Initially it 
has been developed for an application in quality 
assurance in the automobile industry. Due to its open 
system architecture, it can be adapted to a large variety 
of applications using appropriate sensor and light 
configurations in conjunction with a programmable 
software structure. The system includes algorithms for 
digital image processing and multi-image 3D-element 
adjustment. The system concept was first published in 
1990 [Luhmann 1990], and a detailed description of the 
current installation is given in [Loser & Luhmann 1992]. 
A first prototype version was installed at VOLKSWAGEN 
for research and development applications (Fig. 1). The 
initial task was the 3D measurement in a production 
environment of a large variety of second-source parts of 
different sizes, shapes and surface materials. VW's own 
requirements for this system have been changed during 
the period of development, so that the system is currently 
used for internal services and investigations. The 
required accuracy was specified as x0. 1mm (9596) for a 
distance measurement in a measuring volume of 2.0m x 
2.0m x 0.6m. 
Fig. 1. POM prototype, Volkswagen version 
24 
  
1.1 Hardwar 
The VW sys! 
cameras (R 
array Senso 
space of th 
1990]. In a 
cameras for 
mounted in : 
table, which 
devices are 
over 30 light 
(Fig. 2). 
Sensors are 
(Sun Sparc: 
containing | 
control devic 
1.2 Softwar 
The system 
package wf 
processes al 
database 
photogramm 
control and i 
is performec 
with a specie 
The interpr 
statements & 
all processe 
calls require 
network daté 
generated k 
entered by tl 
The photogr 
adjustment [ 
spatial int 
transformatic 
[Andresen « 
method is be