-40-
of each are summarised in Table 1. Scans 1-4, 13-16 and 16-20
were captured with high accuracy and last pulse (standard
parameters) but different resolutions. Four data sets were
captured at each resolution to gauge repeatability, though only
one ultra resolution scan could be captured with available
battery power. For scans 5-8 and 9-12, the pulse mode and
accuracy parameters were respectively changed.
Scan Numbers
Resolution
Accuracy
Pulse
1-4
Coarse
High
Last
5-8
Coarse
High
First
9-12
Coarse
Standard
Last
13-16
Medium
High
Last
17-20
Fine
High
Last
21
Ultra
High
Last
Table 1. Scanner Performance Test Cases.
Figure 1. Histogram of Plane Fit Residuals
(Individual Coarse Resolution, High-Accuracy Scan)
2.3 Experimental Results
Resulting from the testing was a large data set from which
reliable precision estimates could be obtained. In total, nearly
2.1 million individual point measurements were captured. For
each of the 336 (16 x 21) individual scans and the 21 combined
(i.e. mean of 16) scans, a least-squares plane estimate was
removed from the data and statistics compiled from the
residuals. The mean scan was calculated by averaging the 16
sets of co-ordinates for each point. Some of the findings from
these data are presented below.
2.3.1 Individual vs. Mean Scans. For each case, the statistics
of the 16 individual and average (mean) plane fits were
compared. For all high accuracy mode data sets individual scan
precision was ±20-22 mm, while for the mean scan the
precision was ±5-6 mm. Clearly, these follow the basic “one
over the square root of n” rule from statistics. In a more
practical context, this result indicates that a scanner with
seemingly coarse rangefinder precision can yield more precise
measurements through the exploitation of statistics.
2.3.2 Precision vs. Sampling Resolution. Precision was found
to be independent of sampling resolution. Though not
unexpected, this result indicates that one need not necessarily
acquire an ultra resolution scan in order to attain a desired level
of precision.
2.3.3 Precision vs. Pulse Mode Choice. No significant
differences in precision were detected between the first-pulse
and last-pulse data sets. However, it is acknowledged that a
relatively small portion of a wall was imaged at normal
incidence. A more complex shape oriented obliquely may
produce different results. This is the subject of ongoing testing.
2.3.4 Precision vs. Measurement Accuracy Mode. Operation
of the scanner in standard accuracy mode permits measurement
of longer ranges (up to 700 m), which is clearly an
advantageous feature for larger-scale projects or where site
access is restricted. However, the longer range comes at the
cost of lower precision. From four standard accuracy data sets,
individual scan precision was estimated to be ±35-38 mm, while
precision for the mean scans was ±9 mm in all four cases.
Clearly there is a degradation of precision, but in this case only
by about a factor of 1.7 instead of the expected factor of 2
quoted by the vendor.
Figure 1 is a histogram of plane fit residuals from an individual
coarse resolution, high accuracy scan. Figure 2 is a histogram
of residuals for an individual coarse resolution, standard
accuracy scan. The sample size of each was 999 points and
both scans were acquired in last pulse mode. As might be
expected, the high accuracy scan histogram follows a Gaussian
shape. This behaviour was observed for all high accuracy
scans, regardless of resolution or pulse mode choice. Analysis
of the standard accuracy histogram reveals a greater dispersion
and a possible bias indicated by the minor lobe at 0.05 m.
Figure 2 is representative of the other standard accuracy scans.
Whatever the mechanism used for deriving longer-range
measurements, it is clearly biased.
100
90
80
Figure 2. Histogram of Plane Fit Residuals
(Individual Coarse Resolution, Standard-Accuracy Scan)
3. SENSITIVITY TESTING
3.1 Description
As will be described in Section 4, the scanner’s ability to sense
deformation was tested on a real load-testing project.
Laboratory simulations were also conducted to quantify scanner
sensitivity in a controlled environment. The assessment was
performed by analysing differences in surfaces modelled from
laser scanner data rather than on a point-wise basis, as is
commonplace in photogrammetric monitoring campaigns.