The Program PALM for Automatic Line and Surface Measurement
Using Image Matching Techniques
H. Schewe and W. Fórstner
Institute for Photogrammetry
Stuttgart University
0.Introduction
During the last 10 years several algorithms for automatic
Stereopsy have been developed. Theoretical and empirical
investigations have proved that high precision measurements
can be performed with digital image matching techniques.
Though these procedures are not foolproof in an arbitrary
environment they work satisfactory if certain conditions are
met. The smoothness of the surface is the most important
one. A severe restriction also results from the limited size
of digital cameras with usually 512 x 512 picture elements
(max. 2048 x 2048 today). This reduces the relative accuracy
to 10-4 if one assumes a location accuracy of 31/20 of “a
pixel.
Both disadvantages can be circumvented by integrating
digital cameras into theodolites or analytical plotters and
by using the interpretation capability of a human operator
in a semiautomatic system. By exploiting the high relative
accuracy of classical surveying instruments being in the
order of 10-5 to 10-8 as a reference the algorithms for
computational stereo then can be used to replace parts of
the measuring procedure. This essentially results in a
two-level hierarchy of the measuring procedure, the upper,
an operator and hardware level controlling the orientation
or the position of the cameras and the lower, a software
level performing the necessary object recognition and
matching algorithms. This way the human operator still keeps
control of the whole procedure, by at least measuring in
difficult situations. The measuring procedure then possesses
the same accuracy as the reference frame of the surveying
instrument. A smooth interaction between automatic and
manual measurements can always be guaranteed by proper cali-
bration.
This paper wants to present the concept for an automatic
surface measurement procedure as it is realized on a photo-
grammetric measuring device.
The task is to determine the surface of an object at or
in the vicinity of the intersection with a set of predefined
planes of arbitrary orientation. This includes covering,
esp. topographic surfaces in bands with arbitrary or regu-
larily distributed points, the automatic detection of break
- 608 -