The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part BI. Beijing 2008
181
Based on the results of this study, we have concluded that POS
(or any GPS/INS system) data accuracy has the most dominant
impact on the attainable horizontal accuracy of airborne lidar
data. Hence, the specified horizontal accuracy numbers of two
airborne lidar systems equipped with the same or equivalent
GPS/INS systems must be identical, if these numbers are
derived by similar methodologies, and if the same or similar
reference set of operating conditions has been considered.
3.5 Post-Processing and Data Accuracy
Another factor that may have a crucial impact on the accuracy
numbers on the lidar system specification sheet is the data
processing procedure. Even after thorough consideration
including a reference set of operating parameters, a reference
target, and a reference set of data collection conditions, the data
processing procedure and the various processing algorithms
applied to the raw lidar data may introduce or reduce errors.
The data set might be further adjusted, optimized, or smoothed
by using third-party software. Moreover, additional data
optimization algorithms could also be applied to data already
processed and calibrated by the manufacturer’s proprietary
software. After a series of data processing procedures, the final
accuracy numbers may look very different as a result. Since
every commercial lidar system manufacturer uses a unique set
of proprietary procedures to determine data accuracy, there is
always a “grey” area around the accuracy numbers on the lidar
system specification sheet.
Based on the results of a recent study on the impact of the data
processing procedure on lidar accuracy numbers (Pokomy et al.,
2008), we have concluded that the optimization algorithms
applied to the processed lidar data may significantly improve
the derived accuracy numbers. Table 1 and Table 2 show some
results of this study, in which two different software tools and
two different algorithms were used to calculate the RMSE (root
mean square error) and standard deviation for the vertical
accuracy of data collected by three different ALTM systems at
slightly varied flying altitudes of about 1 km and under similar
operational conditions.
Software tool 1
Software tool 2
Algorithm
1
Algorithm
2
Algorithm
1
Algorithm
2
System 1
0.096
0.095
0.087
0.086
System 2
0.053
0.047
0.054
0.047
System 3
0.074
0.054
0.074
0.054
Table 1. Comparison of z-RMSE values for the lidar data
processed by different software tools and different algorithms
Software tool 1
Software tool 2
Algorithm
1
Algorithm
2
Algorithm
1
Algorithm
2
System 1
0.062
0.052
0.061
0.051
System 2
0.043
0.036
0.044
0.035
System 3
0.072
O
Ö
L/1
•-J
0.072
0.057
Table 2. Comparison of standard deviation values for the lidar
data processed by different software tools and different
algorithms
The comparisons in Table 1 and Table 2 show clearly that the
final accuracy numbers presented on the lidar system
specification sheet may be improved by 10-30% simply by
using different processing algorithms, either developed
internally by the lidar system manufacturer or provided by
third-party software.
In addition, since overall lidar data accuracy strongly depends
on the accuracy of the position and orientation data, post
processing software tools available in advanced GPS/INS
systems may also have a significant impact on final data
accuracy. A prime example is the new POSPac 5.0 processing
package offered by Applanix/Trimble, which has proved to be
even more robust than the POSPac 4.4 currently used with
ALTM/Gemini and is capable of handling steeper banking
angles without compromising the specified accuracy (Hutton et
al, 2007). In tests performed at Optech (Boba et al., 2008),
processing with the POSPac 5.0 has consistently shown
improved POS data accuracy that, in turn, improved the overall
accuracy of ALTM/Gemini data.
Thus, the data accuracy derived immediately after data
processing may look noticeably different from the numbers
derived after applying additional processing tools to optimize
the data. Moreover, the methodology that the data processing
workflow uses to derive the accuracy numbers may vary from
one manufacturer to another. Therefore, the final accuracy
numbers derived by different methodologies would not be
obviously valid for sensible comparison.
4. CONCLUSION
To bridge the gap between the numbers on a lidar specification
sheet and expected system performance in the field, the lidar
system user must understand the underlying premises and
relationships between these numbers and plan an airborne
survey project accordingly.
It was shown that in addition to laser PRF, which determines
data collection efficiency, the scan pattern and beam deflection
mechanism used in a particular lidar system may influence
ground point density and area coverage rate and consequently
affect the operating parameters for a planned mission. The
dynamic range of intensities that a particular lidar system can
accommodate may also significantly enhance or reduce
achievable data quality and accuracy. Hence, to collect
accurate data without voids over highly variable terrain, the
user should carefully evaluate a lidar system’s dynamic range
capabilities and limitations.
In addition, it was shown that the combined impact of laser
footprint size and the GPS/INS system on lidar data accuracy
can make the data collected at very high altitudes and very wide
scan angles not usable for most practical applications. Also, the
analysis of the impact of processing algorithms and third-party
software tools on data accuracy indicated that accuracy
numbers derived by different processing workflows may look
noticeably different. Thus, without consensus in the industry
on how to derive the lidar data accuracy numbers, lidar users
should remember that the numbers they see on lidar
specification sheets across different manufacturers may not be
valid for comparison and may not be applicable to certain
survey conditions.
In conclusion, knowing the relationships underlying
manufacturer-derived lidar specifications and the many factors
that can alter actual data collection efficiency and quality will