53 
COMPARATIVE GEOMETRIC TESTS OF INDUSTRIAL AND SCIENTIFIC CCD CAMERAS 
USING PLUMB LINE AND TEST RANGE CALIBRATIONS 
Mark R. Shortis Walter L. Snow and William K. Goad 
Department of Geomatics Experimental Testing and Techniques Division 
The University of Melbourne NASA Langley Research Center 
Parkville, Victoria 3052, AUSTRALIA Hampton, Virginia 23665, U.S.A. 
Telephone : +61 3 344 6806 Telephone : +1 804 864 4613 
Facsimile : +61 3 347 2916 Facsimile : +1 804 864 7607 
Mark_Shortis @ mac.unimelb.edu.au W.L.Snow @larc.nasa.gov 
W.K.Goad@larc.nasa.gov 
KEY WORDS : Calibration, CCD, Plumb line, Test range, Sensor geometry 
ABSTRACT 
Small format, medium resolution CCD cameras are at present widely used for industrial metrology applications. Large 
format, high resolution CCD cameras are principally used for scientific applications, but are slowly infiltrating 
photogrammetry and dramatically improving the object space accuracy achievable by close range measurement. The 
calibration of all types of CCD cameras is necessary in order to characterise the geometry of the sensors and lenses. 
Fourteen different types of CCD sensor and lens combinations have been calibrated at the NASA Langley Research 
Center using plumb line calibration combined with self-calibration and a targeted test field. The results of these 
calibration tests will be described, with particular emphasis on the dependence on sensor resolution and a comparison 
of industrial and scientific CCD cameras. 
INTRODUCTION 
Camera calibration has been and will always be a necessary part of the photogrammetric process. Knowledge of the 
internal geometry of the camera is essential if the principle of collinearity is to be correctly applied. Without this 
knowledge, derived measurements in the object space will be affected by systematic errors and therefore will be 
degraded in accuracy. Whilst in many circumstances self-calibration may be feasible, in some cases such an approach 
is not appropriate. Self-calibration is typically ruled out because the geometry of the photogrammetric network is too 
weak to determine the calibration parameters with confidence or with a reasonable degree of independence. The internal 
characteristics of the camera must therefore be determined before or after the actual measurement process, or 
preferably both. 
Since their introduction in the 1970s, CCD cameras and digital images have gained wide acceptance for machine vision 
and industrial metrology. The very nature of the tasks in which CCD cameras are employed tends to exacerbate rather 
than ameliorate the calibration problem. Real time or near real time applications require multiple, fixed (and 
synchronised) cameras (Childers et al, 1994), whereas self-calibration is most effective with a single, portable camera 
(Fraser and Shortis, 1995). Further, such applications often involve small numbers of targets, whereas a dense, three 
dimensional array of targets which fills the camera format vastly increases the effectiveness of self-calibration. 
Pre- or post-calibration can be carried out by a variety of techniques. Using an established test field, comprising a two 
or three dimensional array of suitable targets with known coordinates, and multiple photographs is perhaps the most 
common method and has been used for virtually all types of cameras (Earls, 1983; Merchant and Tudhope, 1989; Wiley 
and Wong, 1995). However, known target coordinates may not be necessary. If the geometry of the photogrammetric 
network is sufficient (Shortis and Hall, 1989) and only the primary physical calibration parameters (principal point, 
principal distance, lens distortions, image orthogonality and image affinity) are desired then the situation reverts to a 
self-calibration. The calibration is in effect taking advantage of a better network geometry which can not be obtained 
during operational photography. Although it is advisable to include an accurate distance between two targets to 
correctly scale the network, even this minimum information is unnecessary for self-calibration. 
Other techniques: are generally partial calibrations which derive a subset of parameters. The plumb line calibration 
method was originally developed by Brown (1971) for close range cameras, but has been successfully and routinely 
applied to aerial (Hentschel and Shortis, 1991), tube type video (Burner et al, 1985) and CCD video cameras (Fryer and 
Mason, 1989). This technique is capable of deriving the lens distortion parameters independently from all other 
parameters, and can be carried out under operational conditions. Many specialised techniques have been developed to 
empirically or analytically derive particular aspects of the internal geometry of cameras. For example, Lenz (1987) used 
target tracking to determine the principal point location and a 2D point array to determine radial distortion of a zoom lens 
on a CCD camera, the DLT approach to combined exterior and interior orientation of cameras is widely used (Burner et al, 
1985) and simple techniques have been developed for non-specialists to obtain a minimal calibration (Snow et al, 1993). 
CCD CAMERA CALIBRATION AT NASA LANGLEY RESEARCH CENTER 
Close range photogrammetric techniques have been in use at NASA Langley Research Center (LaRC) for almost twenty 
years. Photogrammetric techniques for wind tunnel testing initially used synchronous stereophotography with 
conventional non-metric cameras to measure deformations of wind tunnel models under test conditions (Brooks and 
Beamish, 1977). More recently, state of the art, large format, metric cameras have been routinely used to characterise 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences", Zurich, March 22-24 1995