You are on page 1of 10

Cost effective accurate orthoimaging of inaccessible areas

Panagiotis Fragkos and Charalabos Ioannidis


*
National Technical University of Athens, 9 Iroon Polytechniou St., Athens, Greece
*cioannid@survey.ntua.gr; phone 302107722686; fax 302107722677
ABSTRACT
Although high resolution orthoimages have been the most popular photogrammetric derivative, the cost of orthoimaging
production remains high for areas where access is difficult. This paper presents a procedure for a cost effective
orthoimage production, from data (digital images, GCPs and DTMs) derived solely from Airborne Laser Scanner
systems. ALS systems are designed to produce a highly accurate 3D point cloud, which can be easily rasterized in an
accurate, high resolution and dense DEM. Applying a hillshade effect on that dense DEM, one can visualize the objects
with great detail and thus measure 3D coordinates for points to be used as GCPs. Modern ALS systems incorporate
medium format digital cameras with high resolution imaging abilities. As no ground surveying is needed, the production
of high end orthoimages may be performed even in regions that are inaccessible or difficult to access, such as remote
islets and mountain tops. A workflow is proposed for the production of orthoimages from low density (2.5 points/m
2
)
laser data using an ALS II-50 system and its medium format RCD105 digital camera. The area of the case study is the
northern section of the Greek island Milos. Utilizing automated processes, a series of quality control tasks and a pre-
processing of the data is performed. During the data processing, an interpolation of the LiDAR datas first returns is
performed, in order to rasterize a high resolution (1m) DEM. The orthorectification process is been performed with the
same DEM, leading to an orthoimage with a planimetric accuracy of 1m and minimum geometric distortions. The cost
analysis of the applied procedure has proved that the method is less cost demanding compared to the usual orhoimage
production methods.
Keywords: LiDAR, Orthoimage, DEM, Accuracy, Cost, QC

1. INTRODUCTION
Orthoimages are considered to be the most important basemap in every form of spatial analysis and planning; e.g.,
studying natural phenomena, urban planning, constructions, risk management and assessment, etc. Traditionally
orthoimages have been produced by aerial images combined with ground control point (GCP) measurements and an
appropriate DEM [1], but as modern technologies are continuously emerging, new types of imagery and new sources for
GCPs collection are been used. Since the imagery needed for the orthoimaging process is relatively easy to acquire, the
main issue to be handled is the GCPs and DEM availability. Performing ground measurements is a demanding and
expensive task, due to the fact that the personnel and the equipment needed must work on the field for long periods and
their acquisition speed depends on a variety of factors such as weather, accessibility, GPS baseline lengths needed to be
measured, etc. Moreover, in a lot of cases, collecting GCPs is nearly impossible due to the inaccessible character of
certain remote areas, such as mountain tops and remote islets, which are very common in the Greek landscape. In that
way alternative ways for GCPs collection are needed to be investigated, in order not only to reduce the cost of the whole
procedure but also to allow mapping in those remote areas.
DEM availability is also considered to be a limiting factor in high resolution ortoimaging, since it needs to be derived
photogrammetrically from the stereopairs used for the process. This is due to the fact that free DEMs that can be found
from various sources (e.g., SRTM, GDEM ASTER) are not appropriate in terms of accuracy and resolution for high
resolution orthoimaging. The DEM collection task is time consuming and has a serious demand for expertise, which
overall increases the cost of the production and thus limits the ortoimages acquisition only to those that can afford them.
These limiting factors in the orthoimaging production are overcomed by new spatial data acquisition technologies that
are constantly expand the scientific field of photogrammetry. Mobile Mapping Systems (MMS) where evolved during
the 90s due to the GPS capability in deriving positional data for moving platforms such are aircrafts [2]. Mobile
Mapping Technology (MMT) is the expansion of the term MMS, which was introduced in the middle of the first decade




2

of the new millennium, in order to characterize all the modern technology and the sophisticated methods used for spatial
data acquisition and processing such as LiDAR systems, IfSAR, etc [3].
LiDAR stands for Light Detection and Ranging and is a technology that matured through the 90s, due to the laser
industry expansion, and presents rapid evolvement. A LiDAR system integrates a laser scanner, a positional and
orientation (POS) unit, consisted of a global navigation satellite system (GNSS) and an inertial navigation system (INS),
along with an appropriate management and storage digital unit. The system performs very quick and highly accurate
measurements of distance and angles (polar coordinates) for a large number of ground points (currently up to 500,000),
in a repeated pattern [4], [5]. These measurements, combined with the systems mounting parameters and positional and
ancillary data, produce a 3D point cloud which is actually a dense sampling of the scanned area (up to even 10 or more
points per square meter). LiDAR technology delivers high accurate 3D georeferenced information for the scanned area,
which can be collected day and night and up to current, it could not be acquired in that way. This is the primary reason
for the constantly expanding demand on LiDAR data, for various applications such as 3D city and terrain modeling,
accurate DEM and DTM production, corridor mapping, vegetation biomass estimation, risk management and natural
disaster assessment tasks, homeland security and law enforcement tasks, etc [6], [7]. The current accuracies that can be
achieved are better than 5 cm vertical and 10-100 cm planimetric, and are depending on various factors such as the
performance of the GNSS system, the calibration of the LiDAR, the nature of the scanned area, the flight height, the scan
angle and many more [8]. In addition, LiDAR scanners have the ability to record multiple backscaterings (up to 5) for
the same pulse, which correspond to objects with different height, or different vertical parts of the same object (e.g.,
trees, buildings, etc.) in the scanned area. Usually first returns form the DEM of the area (tree canopy, building surface,
etc), while last returns usually represent the ground, providing the ability to map the ground even in forest areas.
Point clouds derived from modern LiDAR systems can achieve high point densities which leads to a high resolution
dataset of ground measurements, which is significantly denser than any dataset manmade ground measurements can
produce. These point clouds are directly georeferenced and can be rasterized to a series of products (e.g., intensity
images, DEMs, etc.) that inherit the planimetric accuracy of the primary point cloud. Accurate raster basemaps can be
easily produced direct from the scan data, which are actually a limited use equivalent of an orthoimage.
Geocoded point clouds usually are delivered in .las format (binary format established by ASPRS), which contain
information for the point coordinates, GPS times, intensity values and more [9]. Las files are to be further processed,
depending on the task, in order to produce various raster derivatives, as well as classified point clouds where each point
is sorted in a specific class representing ground, buildings, water bodies, vegetation, roads and many more thematic
categories. The classified point clouds can further easily be utilized for feature extraction and accurate area mapping.
Processing LiDAR data can be a challenging task as it requires a lot of computational and storage resources, since every
.las file may contain hundreds of thousands or even millions points, depending on the density of the point cloud and the
size of the scanned area. Moreover, although the processing techniques are automated up to a significant level, there is
still a lot of editing needed especially in the classification process. In the present papers case study and as far as LiDAR
aided ortoimaging is concerned, the main derivative needed is a DEM, both in raster and in elevation model (3d) format,
for the GCP collection and for the differential ortorectification process. This DEM represents the ground surface and the
above ground objects and its production is quite easy and fast contributing to cost effectiveness.
Modern LiDAR systems incorporate a medium format digital camera, with high resolution capabilities, thanks to the
advance in digital systems and optics of the past decades. This frame camera covers the scanned area with digital
overlapping images, at the same time the scan is been performed, providing the necessary data in order to produce
orthoimages.
2. METHODOLOGY
The proposed methodology in this paper includes preprocessing and processing stages for the preparation of the LiDAR
data and the DEM extraction required for the GCP collection and the orthorectification. The overall workflow is been
presented in Figure 1.
Initially, a spatial subset of the LiDAR data is created, in order to limit the need for computational resources, by reducing
the scanned area size, to match with the size of the area to be orthoimaged. This step is only necessary in cases where the
scanned area exceeds the size of the area of interest.





3


Figure 1. Workflow diagram
In the first stage, a series of quality control checks are been performed, in order to validate the internal accuracy (IQC) of
the LiDAR data. Internal accuracy depends on the datas completeness degree, in terms of area coverage, measurement
consistency, density, etc. Visual check is been performed with the aid of 3D viewers, that LiDAR processing software
incorporate. Visualizing the point cloud is the only way to have an original assessment of the objects that can be
distinguished over the scanned area, and thus to make an initial check for the potential GCP availability. Also,
visualizing the data reveals instantly any discontinuousness as far as the scan coverage concerns. The scan footprints are
been calculated from the data itself in order to check the area coverage in a more precise way, and by overlay in a GIS
the data consistency is measured and estimated. Next, the point clouds statistics are been calculated, which include mean
density and nominal spacing values; number of points that are first to last returns; min and max elevations measured, as a
check for non-acceptable measurements. Elevation values higher than the local maximum correspond to backscatters
from flying objects like birds, while measurements with negative elevations are usually occur when a water body is been
scanned, due to the laser pulses limited penetration in it. Finally, an estimation of the inerstrip vertical accuracy is been
held, by forming dZ images in the overlapping areas. Those images visualize the distribution of the elevation differences
which occur when the same area is been scanned by two adjacent flight lines and may result to unacceptable elevation
variations if large systematic errors are present.
Having assured that the LiDAR data are qualified for the tasks accuracy standards, a preprocessing stage is performed
where the data are been prepared for the main process of the DEM extraction. In this stage, a reprojection of the data is
been performed if necessary, and the separate flight lines are been merged to a single file. Points with non-acceptable
elevation values are filtered from the point cloud and the scan data are distributed in custom sized tiles. Tiling the
LiDAR data is a standard technique for processing smaller files, reducing the computational resources needed due to the
large number of points.
In the main processing stage, the point cloud is rasterized according to the intensity value that has been recorded for each
measured point. The intensity image that is extracted, can be used as a basemap for performing planimetric checks,
and is a georeferenced raster similar to a panchromatic image, which inherits the plannimetric accuracy of the point
cloud. Finally, the DEM of the area is been interpolated and a model elevation calibration is been performed, in order to
compensate for the local geoid height, transforming the ellipsoidal heights of the LiDAR to orthometric heights.




4

In the final stage, the DEM is been visualized in a hillshade effect and then utilized to identify and collect GCPs for the
orthoimaging process. Using the medium format digital camera images of the LiDAR system, an aerial triangulation with
bundle block adjustment method is been calculated and the differential orthorectification process is been performed
using the LiDAR derived DEM, with no additional breakline collection. The orthoimage is been examined to evaluate its
planimetric accuracy and its optical quality (existence of optical distortions, etc).
3. APPLICATION
3.1 Area of study and collected datasets
An application of the proposed methodology was conducted over an area in the Greek island of Milos, which is located
in the Cyclades island complex in the Aegean Sea. A LiDAR scan along with digital image capture, performed by the
UK National Research Environmental Council (NERC). The scan covered the whole island in 29 overlapping strips and
2 vertical calibration strips. A small area of 5.7 km
2
in the northern part of the island, covered by four (4) digital images,
was chosen as a case study (Figure 2). Minimum orthometric elevation for the study area is 0 (sea level) and maximum
elevation is 254 m. The LiDAR system used was a Leica ALS50-II with an incorporated Leica RCD105 medium format
digital frame camera. Scan parameters and cameras calibrated values presented on Table 1.


Figure 2. Scan data and imagery coverage over Milos Island (left) and the selected area of study (right)

Table 1. Scan parameters and RCD105 camera parameters
Scan parameters RCD105 camera parameters
Pulse repetition frequency (PRF) 99900 KHz Resolution 7216 x 5412 pixels (39 Mega-pixels)
Scan frequency (fsc) 52.8 Hz Calibrated resolution 7212 x 5408 pixels
FOV 18
o
Calibrated CCD dimensions 49.0416 mm x 36.7744 mm
Flight speed 140 knots Calibrated pixel size 6.8 m
Flight height (AGL) 2500 m Calibrated focal length (c) 59.799 mm
Percentage of returns
Min: 32.1 %
Max: 100.0 %
Mean: 67.2%
Calibrated lens distortion
coefficients
k0= 8,57325 x 10
-3

k1= -2,01969 x 10
-8
k2= 5,13135 x 10
-11

Mean GPS PDOP 1.4 Calibrated primary point pos. xo= -0.0025 mm yo= -0.3247 mm
Mean GPS satellites observed 14 Exposure time Up to 1/4000

3.2 Quality control
Performing the quality control procedures over the LiDAR data, we visualize the point cloud with ERDAS 3D viewer,
colored by intensity values (Figure 3). From the 3D model we can distinguish low trees and buildings, as well as linear
features on the surface. Also, the area coverage is checked for inconsistencies, by extracting the scan lines boundaries in
shp format with LasTools and overlaying them in the ArcGIS. The output shows perfect coverage for the area of interest.
Furthermore the statistics for the point cloud are calculated (Table 3):
- An average density of 2.5 points/m
2
is extracted, classifying the point cloud as low density.
- The nominal spacing between points is 0.63 m and the total number of points in the area is approximately 14.6
million points. The vast majority of the points measured are first returns, which indicates that the coverage of the area
(high vegetation, buildings, etc.) is relatively sparse, since not many multiple backscatters where caused.




5

- The maximum (geometric) elevation value of 885.47 m is not acceptable. The maximum orthometric elevation of the
area is 254 m, as derived from an existing local topographic map. An approaching value of less than +40 m for the
geoid height of the area is calculated using EGM08. So, no point in the area should be elevated more than 294 m with
respect to ellipsoid. Furthermore, the lowest ellipsoidal elevations in the area should present a value of approximate
40 m, concerning the sea in the scanned area. The minimum value of 2.28 m in the point clouds geometric elevations
is obviously due to the infiltration of the laser in the sea. Points with non-acceptable elevations will be filtered in the
preprocessing stage using appropriate boundaries.

Figure 3. 3D visualization of the point cloud over the area
Table 2. Statistics of the point cloud
Parameter Value Parameter Value
Total point number 14,644,074 Mean point density 2.5 points/m
2

1
st
Returns 14,525,127 Min 2.28 m
2
st
Returns 118,352 Max 885.47 m
3
st
Returns 590 Area size 5.7 km
2

4
st
Returns 5 Number of unclassified points (class 1) 14,644,074
Nominal point spacing 0.6 m

Checking the LiDAR data for systematic errors, elevation difference (dZ) images in the overlapping sections of the scan
lines are produced. For every overlapping section, a 1 m grid is been formed, since nominal point spacing is 0.63 m,
ensuring that at least one LiDAR point falls inside every pixel (1x1 m). For every pixel in that grid, two elevations are
calculated, one from each adjacent scan line. Those elevation values are the mean elevation of all LiDAR points that fall
inside that pixel. The difference of those elevations, is the dZ value that is been assigned as a pixel value for the image.
Then, this dZ value is been compared by a user threshold -n- and the pixels of the dZ image are colorized according to a
pallet, visualizing the elevation differences. Using the LasTools palette the pixels are colorized from blue to white to red,
corresponding to elevation differences from less than n to 0 to more than +n. Performing the check three different
thresholds of 1 m, 0.5 m and 0.25 m are used. The corresponding dZ images showed low concentration and no
systematic distribution of elevation differences (Figure 4).

Figure 4. Elevation difference (dZ) images with 1 m (left), 0.5 m (center) and 0.25 m (right) threshold values.

3.3 Preprocessing LiDAR data
In the preprocessing stage, the projection of the LiDAR data is been defined in the header of the las files. Since the data
of the this application where third party data and no information for the coordinate system was at present, a coordinate
check applied, showing that the point cloud was referring to CGS WGS84_UTM_Z35N. The used coordinate system
was defined in the header and the data reprojected to CGS GGRS87_TM projection (Greek Grid). The separate las files,
which represent each flight line, are merged into one single file which contains the exact same records as the original




6

files. The unified file is relatively difficult to process, due to its size and is been utilized only for non-acceptable
measurements filtering and for the tiling process. Filtering elevations accordingly to the acceptable bounds defined in the
quality control stage, we process the point cloud in LasTools, keeping only points with elevation that lies within the [35
m, 300 m] space. The bounds used for the elevation filtration are expanded for safety reasons and they do not affect the
outcome, since any points that may lie within these expansions will be very few in numbers with respect to the whole
dataset.
Proccedingly, a tiling is been applied in the merged point cloud, in order to create smaller files with specific NxN size,
that correspond in the tasks demands and will be able to be processed according to the computational resources of the
system used. 34 tiles at a size of 500x500 m are created, with a 50 m overlap (buffer around every tile) in order to be
able to cut off any artifacts appearing in the tile edges after processing.

3.4 DEM extraction
Once preprocessing stage is complete, the LiDAR data can be exploited for derivative production. The point cloud is
rasterized based on the normalized intensity values recorded by the sensor for every point, in order to create a basemap
for comparison and check (Figure 5). The intensity image is been interpolated with a spatial resolution of 1 m, due to
the point clouds nominal point spacing of 0.63 m, which ensure that every pixel of the image acquires value by at least
one point that falls within its boundaries. The 2.5 point/m
2
low density point cloud produces a high resolution but low
sharpness intensity image, with a relative fuzziness around the various objects edges. Clearly, the higher the density of
the point cloud, the higher the intensity images quality will be.

Figure 5. Intensity image of 1 m resolution; right: detail of the image
The interpolation of the DEM takes place by TIN process over the total of the points that classify as first returns.
Choosing only the first backscaterrings of the laser pulses emitted, we ensure that only points which describe the scanned
surface (ground along with the supernatants objects) will be processed. A spatial resolution of 1 m is been chosen (as for
the intensity image) and a grid is formed. By Delaunay triangulation, a TIN is been calculated for the total of first return
LiDAR points, and every pixel of the grid acquires an elevation value by interpolation on the corresponding TIN
triangle. By applying a hillshade effect along with an intense color palette, the objects in the study area become discrete
and measurable (Figure 6).

Figure 6. DEM (left) and hillshaded DEM (right)




7

Since no accurate geoid height has been calculated for the area the transformation of the geometric heights to orthometric
elevations are made using the elevation measurements in the sea area, which are approximately 39.9 m. Any deviations
from the true geoid height, which refers to the mean sea level, with tidal corrections, will cause a systematic bias in the
whole model that can be easily eliminated at any stage.
Having available a set of known points with orthometric heights that have been measured over the area, an elevation
accuracy test is performed. By overlaying these known points, a comparison is been made between the known elevations
and the calculated ones by the model, in the specific locations. Test shows a RMSE of 1.56 m, which is almost
systematic, since the calculated deviation of the mean height translation is 0.18 m (Table 3). The accuracy of the LiDAR
derived DEM appears to be high as far as elevations concern, and by translating the model -1.55 m, a vertical accuracy of
almost 20 cm is achieved. The major part of this remaining 1.55 m bias is obviously a result of the non-accurate geoid
height that was used for the model calibration.
Table 3: DEM elevation check

Measured elevations
Zm (m)
Elevations by DEM
interpolation ZDEM (m)
(m)
(m)
H
(m)
RMSE
(m)
1 36.28 34.60 -1.68
-1.55 0.18 1.56
2 47.13 45.25 -1.48
3 20.53 203.92 -1.61
4 132.88 130.95 -1.33
5 253.54 251.85 -1.69
6 147.09 145.31 -1.77
7 142.94 140.94 -1.30

A fast planimetric check, by overlaying the hillshaded DEM (without the color palette) over the intensity image, shows a
perfect match (Figure 7). This is to be expected, as the only procedure that comes between the point cloud and the raster
derivatives is a simple interpolation, which does not result in important planimetric errors.


Figure 7. Intensity image (up) and hillshaded DEM (bottom)

3.5 Orthoimaging process
Evaluating the LiDAR systems imagery and given the parameters of the flight height H=2500 m, the RCD cameras
calibrated c=59.799 mm and the image pixel size of 6.8 m (Table 1), the calculated groundel equals to 0.29 m and the
suitable pixel size of the orthoimagery should be smaller than 0.50 m, which is a resolution capable for printing
orthoimages up to a scale of 1:5000 according to the specifications of the Greek National Cadastre and Mapping Agency




8

(NCMA). In addition, an appropriate DEM for the orthorectification process should at least have a resolution of 7.5 m,
which is far lower than the 1 m resolution DEM derived from the LiDAR data.
The GCPs coordinates uncertainty, the point image coordinate measurement errors during the aerial triangulation
process and the vertical accuracy of the DEM are the main sources of the planimetric error in the final orthoimage, which
should be less than 1.25 m. Given the very accurate and very dense DEM and by performing careful point measurements,
it is fairly safe to assume that the main error factor will be the coordinate error of the GCPs. An a-priori estimation, is
that by performing GCP measurements on the DEM, we will accept at least 1 pixel error. Given the 1 m DEM resolution,
the corresponding planimetric error for the measured GCPs is 1 m at best, which is considered to be on the edge of
acceptance.
The hillshaded DEM is best utilized with an intense color palette, for best object discrimination. The edges of the
buildings appear pixelated, but there is a variety of characteristic objects that can be utilized as GCPs. As shown on
Figure 8, small trees and bushes, surface anomalies, dry stone wall junctions (which are very common even in the remote
Greek island landscape), road junctions and many more are well discriminated over the area, producing a rich GCP
measuring field. A major advantage in this method is the ability to ideally choose the GCPs locations, so they are optimal
distributed over the area.








Figure 8. GCP distribution (left) and an example of GCP measurements over DEM (right top) and imagery (right bottom)
For the aerial triangulation process, a total of 12 GCPs were measured and a bundle block adjustment was performed
with ERDAS LPS. Results are:
RMSE (X) = 0.53 m, RMSE (Y) = 0.65 m, RMSE (Z) = 0.81 m and RMSE (XYZ) = 1.17 m or 1.87 pixel (< 2 pixel).
The orthorectification process is been conducted with the utilization of the LiDAR derived DEM, without the need to
collect additional breaklines. This is due to the fact that the LiDAR derived DEM is very dense, giving an accurate
description of the surface, which actually over qualifies for the scale of 1:5000. Accurate elevations for inverse
collinearity equation calculations are ensured by the nature of the process followed. Figure 9 presents the final
orthoimage of the study area.




9


Figure 9. Orthoimage of the study area
3.6 Quality control
In order to evaluate the quality of the derived orthoimage, an optical control for geometric distortions and a planimetric
control are performed. Optical distortions in the orthoimage, that would be a result of non-accurate elevations, are
practically nonexistent in the image (Figure 10), which proves the benefits from the as is use of the LiDAR derived
DEM in the orthorectification process. The planimetric check was performed by comparing the orthoimage with the
hillshaded DEM, showing a translation of about 1m RMSE, fairly expected due to the GCP coordinates uncertainty
(Figure 9), and qualifying the image for utilization in a scale of 1:5000.

Figure 10. Orthoimage geometric distortion (left) and planimetric check (right)
4. CONTRIBUTION TO COST EFFECTIVENESS - CONCLUSIONS
The need for cost effectiveness that characterizes our postmodern times, due to factors such as limited funding, energy
saving and environmental protection demands, deliver a major challenge in modern engineering. A viable way of doing
things must have the ability to deliver the best outcome with the least use of the available resources. Modern technology
evolves rapidly, giving solutions that may exploited in various ways and combinations, in order to keep producing, but in
more flexible ways than the traditional ones.




10

In the present paper, a method for producing the basemaps needed (at a scale of 1:5000) for all kinds of spatial planning
and mapping projects is demonstrated, by utilizing solely a complete LiDAR system (ALS) and a derived low density
point cloud from it. It is obvious that with denser point clouds achieved accuracies can be higher, due to the improved
resolution that the derivative rasters will have. The system can deliver both imagery and elevation data in the project
area, eliminating the need for acquiring or producing supplementary ones. The specific process has various advantages
and result to the overall project economy. The need for expensive field surveys, in terms of personnel reimbursement,
surveying equipment cost and time consumption, is practically eliminated, giving a competitive advantage towards
classic procedure production.
The proposed method focuses in an alternative way for the collection of the GCPs and the DEM production. Practically,
the orthoimage production procedure needs tie points between the imagery and the LiDAR data, instead of using GCPs
(and their expensive field measurements). So, the procedure may be easily incorporated in existing photogrammetric
production lines of various agencies and companies worldwide. A major issue in assimilating new technologies and
practices is always the cost of the inevitable production slowdown, which, in the case present, is not an issue, since the
need for new software is minimum, as well as the need for personnel retraining. Classic photogrammetric procedures are
still followed, but enriched in a costless way.
The data derived from the LiDAR system, even though they might utilized for performing orthoimaging processes,
remain suitable for other tasks that present a demand for elevation data. With techniques such as Lidargrammetry, one
can produce 3D vector datasets, by using classic stereo digitizing procedures, over intensity stereomates that may derive
from a single high density point cloud. Being able to use the same datasets, in order to complete multiple tasks, as well
as to produce other datasets from those, is a major contribution to cost effectiveness. IfSAR technology stands as an
alternative for producing accurate high resolution DEMs (applies to airborne SAR systems, since spaceborn SAR are not
comparable to airborne LiDAR, in terms of accuracy) [10], but there is no produced point cloud for the Interfereometric
radar images, while their acquisition cost remains relatively high. Furthermore, though Radargrammetry can deliver
similar results as Lidargrammetry, there can be no accurate feature classification and extraction with the aid of
automated processes that currently apply in the point cloud processing.
Finally, in the case of remote areas where accessibility is limited or even restricted by terrain morphology or sailing
conditions, the presented approach stands as one of the few options for mapping or monitoring. Especially in landscapes
similar to the Greek one, where high and dense vegetation is relatively limited, a LiDAR scan will certain serve the
purpose of the identification and collection of the GCPs. This applies also in cases of natural disaster circumstances,
e.g., earthquakes, volcano eruptions, floods, fires, etc, where field surveys on site may present high risk of human losses,
but the need for a fast, cost effective and accurate assessment is still a demand.
REFERENCES
[1] Kraus, K., [Photogrammetrie] Ferd. Duemmler Verlag, Bonn, 332-340 (2010).
[2] Novak, K., "Data collection for multi-media GIS using mobile mapping systems," GIM 7(3), 30-32 (1993).
[3] Schwarz, K.P. and El-Sheimy, N., [Digital Mobile Mapping Systems State-of-the-Art and Future Trends,
Advances in Mobile Mapping Technology] Eds. Tao and Li, Taylor & Francis, 3-18 (2007).
[4] Pack, R. T., Brooks, V., Young, J., Vilaa, N., Vatslid, S., Rindle, P., Kurz, S., Parrish, Christopher E., Craig, R.,
and Smith, P.W., [Manual of airborne topographic LiDAR Chapter 2: An overview of ALS technology] Renslow
M.S. (editor), ASPRS, Maryland, 7-97 (2012).
[5] Toth, C.K., "R&D of mobile LiDAR mapping and future trends," Proc. ASPRS Annual Conference, Baltimore
(2009).
[6] Deng, F., Zhang, Z. and Zhang, J., "Construction 3D Urban Model from LiDAR and Image Sequence," The
International Archives of Photogrammetry, Remote sensing and Spatial Information Sciences, vol. 34 (2007).
[7] Wechsler, S. P., [Development of a LiDAR Derived Digital Elevation Model (DEM) as Input to a METRANS
Geographic Information System (GIS)] California State University, Long Beach (2011).
[8] Baltsavias, E.P., "Airborne Laser Scanning: basic relations and formulas," ISPRS Journal of Photogrammetry &
Remote Sensing 54(2-3), 199-214 (1999).
[9] Lewis, G., [Topographic Laser Ranging and Scanning Chapter: Management of LiDAR data], Shan, J., Toth, C.
(editors), CRC Press, 295-306 (2009).
[10] Mercer, B., "Comparing LIDAR and IFSAR: What can you expect?," Proc. Photogrammetric Week, Stuttgart
(2001).

You might also like