60
SYSTEMATIC POINTING ERRORS WITH RETRO-REFLECTIVE TARGETS
Roland Zumbrunn
LEICA AG; Moenchmattweg 35, CH-5035 Unterentfelden.
Phone +41 64 45 68 18, Fax +41 64 43 07 34
KEY WORDS: Pointing errors, Retro-targets, Videogrammetry.
ABSTRACT
Accurate target detection is fundamental to many industrial measuring applications. One type of target uses retro-
reflective material for defining object surface points. In a photogrammetric procedure cameras at different positions
illuminate these targets which are imaged onto the CCD array. By intersecting all the lines of sight, the targets are fixed
in space.
In the past accuracy improvements concentrated mainly on algorithms or sensor calibration, such as better modelling
of optical and electronic distortions, and less effort was spent on potential targeting problems. It is conventionally
assumed that the target centre is physically fixed and independent of the angle o. between the target normal and the
line of sight. The following discussion will investigate this assumption in the case of retro-reflective targets.
Tests have revealed a systematic shift of the target center from - 30um to + 30um normal to the line of sight, when the
target is viewed from different directions, i.e. - 40° < a < + 40°. There is no shift seen for punched retro-reflective
targets.
1. INTRODUCTION
Non-contact three dimensional coordinate determination of object surface points by optical means is an important field.
Different optical methods have been developed for dimensional surface analysis of objects of any size. In particular the
method of Close Range Photogrammetry for industrial measuring applications is now well accepted (Wester-
Ebbinghaus, 1989). In this method, object points are marked by different types of targets. These targets are imaged
onto one or more calibrated and oriented CCD cameras placed around the object. For high contrast, targets made of
retro-reflective material and illuminated from the camera are often used. Such targets are available from a number of
suppliers, e.g. Hubbs Machine & Manufacturing Inc..
The circular "retro-targets", 2 to 10 mm in diameter, consist of small glass spheres with an index of refraction n = 1.8.
The spheres (e.g. © = 50um) are glued onto a paper base. On the other side a circular light absorbing mask defines
the edge (contour). This mask is created either by a layer of black paint less than Sum thick or a thin black paper sheet
approximately 20um thick. Along the edge spheres can be partially occluded. When illuminated from the camera
position, a well defined target contour of high contrast can be observed. There are different interactions between the
incident light and the polished glass spheres. Part of the light is reflected at the surface, producing a small spot in the
center of each sphere. A larger part is totally reflected inside the spheres and seen from a distance as a concentric ring
of light. The ring’s diameter is approximately 80% of the sphere’s diameter, depending on the index of refraction. Only
a few percent of the incident light is reflected back along the incident direction, the rest is scattered.
The target is therefore defined by a cluster of many small individual light spots. Depending on the optical magnification,
either a homogeneous light distribution is imaged or a collection of single light spots is resolved by the CCD sensor.
This influences the precision with which the target can be located.
Careful observation of the partially occluded edge spots have revealed a very systenaa behaviour. As the angle
between the line of sight and the target normal increases from zero to around 40°, the intensity of spots on the near
edge decrease and on the far edge increase. In extreme cases new spots may be formed or existing ones may
completely vanish. The result is a shifting target because the effects on the near and far sides do not cancel each other
out.
An estimate of this shift for targets with painted edge masks will be derived, firstly when viewed with a microscope and
secondly when viewed with a video-theodolite.
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences", Zurich, March 22-24 1995
PN FA 1 "V