Geoscience Reference
In-Depth Information
where,
X = Parameter to be evaluated.
r
X
= Estimated standard deviation of parameter.
The hypothesis of the test is:
H
0
: The parameter is not significant to the scanner observation.
H
A
: The parameter is significant to the scanner observation.
The null hypothesis (H
0
) is rejected if the calculated t value (Eq.
12
) is higher
than the critical t value (predicted from the t-distribution table based on the value
of degree of freedom) with selected level of significant (confidence level 95 %
equal to 0.05 of significant level). With the rejection of H
0
, the test parameters is
statistically significant (accept H
A
).
5 Methodology
5.1 Preparation of Test Points
As illustrated in Fig.
2
, industrial or close range photogrammetry can provide sub-
millimetre accuracy. Thus, this measurement technique was used to establish 15
test points (Fig.
8
) which subsequently were used to investigate the capability of a
self-calibration to improve the quality of TLS data.
In order to perform photogrammetric measurement, digital camera Sony DSC
F828 with pixel count 1920 9 1080 (8 megapixels) and 0.0027 mm pixel size was
employed to capture the images of test points. As a routine procedure, the camera
should be calibrated before it can be used for 3D measurement purposes. Figure
9
shows the calibration procedure carried out for digital camera Sony DSC F828 and
the processing of the calibration parameters was made by using Photomodeler
V5.0 software. Several scale bars were positioned at the measurement field (Fig.
9
)
for the purpose to set a scale value as well as to evaluate the accuracy of the
calculated 3D coordinates of test points.
Fourteen independent vectors were generated from these test points to ensure
that it can be used to evaluate the accuracy of TLS raw data and calibrated data
(Fig.
10
). By computing the discrepancies of the benchmark (vectors obtained
from photogrammetry) and the vectors yielded from TLS raw data and calibrated
data, then the accuracies of both TLS data can be statistically calculated. After-
ward, those values were used to evaluate the significant of self-calibration thus
improving the accuracy of TLS data.
