Improvement in Accuracy Through Self-Calibration for Panoramic Scanner - Geoinformation for Informed Decisions - page 200

Fig. 7 Simulation residuals for panoramic scanner, constant rangefinder offset error (a),

collimation axis error (b), trunnion axis error (c) and vertical circle index errors (d)

adjustment that excludes the relevant calibration parameters. Most of the cases, the
trend of un-modelled systematic errors closely resembles the analytical form of the
corresponding correction models. As depicted in Fig. 7 , all systematic error
models are identified by plotting a graph of adjusted observations against resid-
uals. The graph of adjusted range against its residuals indicates a constant
rangefinder offset error (a 0 ) if the trend seems like an inclining line (Fig. 7 a).
When residuals of horizontal observations against adjusted vertical angles shows
the trend like secant function, which means that the scanner has collimation axis
error (Fig. 7 b). Trunnion axis error can be identified by having a trend like tangent
function as shown in Fig. 7 c. For vertical circle index error, by plotting graph of
adjusted vertical angles against its residuals, this systematic error model considers
exist when the trend looks like the big curve as depicted in Fig. 7 d.
Nonetheless, it is quite difficult to identify these systematic errors based on
residual pattern graph especially when the magnitudes of the errors are very small.
As a result, this study has also implemented statistical tests to determine the
significant of the parameter to the scanner observations. Known as t-test, the
analysis is carried out using formula (Gopal 1999 ):

t ¼ X
r X

ð 12 Þ