See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/32894099

Integration of LiDAR and IFSAR for mapping

Article · January 2004

Source: OAI

CITATIONS READS

22 934

1 author:

Ian J. Dowman
University College London
147 PUBLICATIONS   1,634 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A Graph-based Technique for Analysis and Visualisation of Higher Order Urban Topology View project

All content following this page was uploaded by Ian J. Dowman on 15 September 2014.

The user has requested enhancement of the downloaded file.

 INTEGRATION OF LIDAR AND IFSAR FOR MAPPING

Ian Dowman
University College London, Gower Street, London, WC1E 6BT, UK
idowman@ge.ucl.ac.uk

Invited paper Commission II, WGII/2

KEYWORDS: LIDAR, Interferometric SAR, Integration, Mapping, DEM

ABSTRACT

LiDAR and IfSAR data is now widely used for a number of applications, particularly those needing a digital elevation model. The
data is often complementary to other data such as aerial imagery and high resolution satellite data. This paper will review the current
data sources and the products and then look at the ways in which the data can be integrated for particular applications. The main
platforms for LiDAR are either helicopter or fixed wing aircraft, often operating at low altitudes, a digital camera is frequently
included on the platform, there is an interest in using other sensors such as 3 line cameras of hyperspectral scanners. IfSAR is used
from satellite platforms, or from aircraft, the latter are more compatible with LiDAR for integration. The paper will examine the
advantages and disadvantages of LiDAR and IfSAR for DEM generation and discuss the issues which still need to be dealt with.
Examples of applications will be given and particularly those involving the integration of different types of data. Examples will be
given from various sources and future trends examined.

INTRODUCTION of the pulse might partly penetrate the tree canopy and/or travel
through and reach the ground as it can be seen in figure 1. In
Geospatial databases are becoming increasingly important in that case, the sensor will also record intermediate returns when
many areas. There is an increasing demand for National the pulse hits various parts of the canopy and the last return, the
Mapping Agencies to provide geospatial data – to be used by return from the ground.
utility companies, environmental agencies, transport agencies
and industry such as telecoms. At the same time mapping
organisations are looking to use new technology to satisfy these
requirements. Two of these important new sources are LiDAR
and IfSAR, acquired from airborne and spaceborne platforms.
Data from these sensors has been applied to a number of novel
applications such as mapping flood plains, powerlines and
transport infrastructure. This paper sets out to define the role of
photogrammetry and remote sensing in this, and. in particular,
the role of IfSAR and LIDAR.

The paper will first set out the characteristics of the sensors and
the data, and the products being generated. It will then deal
with airborne data collection also look at the data producers
and discuss some of the open questions relating to the use of
LiDAR and IfSAR. Some characteristics and aspects of
spaceborne IfSAR will be considered. Finally the paper will
look at how the technology and applications are progressing.

2. THE CHARACTERISTICS OF LiDAR AND IfSAR

2.1 LiDAR
Figure 1. Illustration of Airborne LiDAR
The principles of LiDAR are well known. These have been
described by Baltsavias (1999b). To summarise: range is Not all laser scanners collect multiple returns, in many cases
measured from a platform with a position and attitude only single returns (first or last pulse) are recorded. In addition,
determined from GPS/INS using a scanning device which most systems record the reflected intensity image.
determines the distance from the sensor to the ground of a series
of points roughly perpendicular to the direction of flight. Figure Lasers also operate as continuous wave sensors which can
1 shows schematically a laser scanner and its main components. depict the interaction between the laser energy and the elements
As a result, the raw airborne LiDAR data is collected in the of the vegetation canopy.
GPS reference system WGS 84.
There are currently many LiDAR systems available and these
The wavelength in which most lasers operate is in the range of operate from fixed wing and helicopter airborne platforms at
1040-1060 nm. (Baltsavias, 1999a) Airborne laser scanners can altitudes from 50 – 3500m. The latest systems operate at
record up to 5 different returns (multiple returns). If a laser 100Hz and can produce point densities from helicopters of 30
pulse or a part of the pulse is reflected from a roof top or the top points per m2. There are now many companies operating LiDAR
of a tree, the sensor will record the first return. However, a part systems, most of which work commercially to provide data to
order, using systems such as Optech, TopEye, FLI-MAP, The two scenes may be obtained from a repeat pass, usually
TopoSys, TerraPoint and Leica. from satellites, where the images are acquired from two passes
of the sensor in very similar orbits. Single pass data is acquired
Spaceborne LiDARs are also in operation, the most important of from an aircraft or spacecraft on which there are two antenna
these is the Geoscience Laser Altimeter System (GLAS) on separated by a known base length. The suitability of a pair of
ICESat. GLAS produces a series of approximately 70 m images for generating IfSAR DEMs is measured by the
diameter spots that are separated by nearly 170 m along track. coherence between them. Poor coherence is caused if the
returned radar signals are different on the two images, phase
2.2 IfSAR unwrapping cannot then be carried out. Coherence is usually
good on single pass images but can be poor on repeat pass,
Synthetic Aperture radar determines the amount of scattered especially if there is a significant time difference between
energy returned to the antenna, its range and position along images. Errors due to the atmosphere are also reduced.
track (azimuth). SAR can operate in a number of frequencies
shown in table 1. The elevation measured for any pixel (resolution cell) results
from a combined signal of scatterers located in the resolution
Band Wave length Frequency cell (sample area). Elevations measure the ‘volume scatter’, i.e.
X 3cm 9.6GHz there will be some penetration into the canopy and the range
C 5.3cm 5.6GHz recorded will not depict the true height of the tree (first surface).
L 24cm 1.3GHz Therefore, areas covered by vegetation will include more height
P 68cm 0.3GHz measurement noise than areas covered by specular scatters (i.e.
buildings). The wavelength of the radar will determine the
Table 1. Typical wave length and frequency for SAR bands. penetration on the signal into the vegetation, X band will not
penetrate as far as L band.
Two SAR images can be combined to use the technique of
interferometric SAR (IfSAR) to generate digital elevation In addition, the surface area represented by one pixel may
models. The principle of IfSAR is shown in figure 2. consist of a combination of different scatterers. Height
measurements could be biased due to a interaction of these
surface features. The backscattered signal (radar response) is
A2
B integrated over a square footprint (resolution cell) somewhat
A1 α larger (about 50%) than the 5m DSM sample distance. (Mercer,
2002) Therefore, the elevation measured for any DSM sample
(resolution cell) will result from a combined signal of scattering
p + δp objects located in this sample area. If hedges and shrubs are
p
closely located to a road, both, the raised objects and the road
θ
itself (bald earth) will contribute to the elevation value
H measured for this DSM sample.

IfSAR has been widely used from spaceborne platforms, the

ERS Tandem mission and the Shuttle Radar Topography
Mission (SRTM) are the two prime examples. The main
h airborne IfSAR is the Intermap STAR-3i. This is a single-pass
across-track IfSAR system operating commercially since
January 1997. The system is an X-band SAR interferometer
Figure 2. Geometry of single-pass Interferometric SAR carried on board a LearJet 36. The two antennae are separated
by a 1m baseline. Accurate positioning and orientation is
The two antennae are shown at A1 and A2. H represents the achieved through the use of an on-board laser-based inertial
altitude above the reference ellipsoid, h indicates the navigation system and an on-board differential GPS (Global
topography of the Earth’s surface. The baseline, i.e. the Positioning Systems) system. (Mercer & Schnick, 1999). Other
separation between antenna 1 and 2 , is given by B. The slant airborne SARs are operated by research organisations such as
range (look direction of the antenna) to the target is given by p, NASA and DLR.
the look angle at target location is represented by and the
angle of the baseline with respect to the horizontal is given by . 2.3 Products and data providers
Assuming is known, the elevation of the targeted point on the
Earth’s surface can be calculated from: Data from LiDAR or IfSAR is usually provided as digital
surface models (DSM), digital terrain models (DTM) and
h = H − p cos θ orthoimages. The generation of the DSM will be done by the
( p + δp ) 2 = p 2 + B 2 − 2 pB cos(90 − θ + α ) organisation which has collected the data and will involve
calculation of the ground co-ordinates from the GPS, INS and
range measurement, and must include corrections derived from
Where δp is the slant range difference the system calibration and from the atmosphere, and of course
δp = λφ / 2π be delivered on a specified datum in a known map projection.
Where is the wavelength of the radar pulses and the phase An image can be formed from the SAR data and may also be
difference between the two returns. The phase difference can be collected with LiDAR. The DTM and orthoimages can be
measured only as a variable with 2 period. Therefore, phase derived from the DSM and image; this will be discussed in
unwrapping needs to be applied in order to resolve the absolute section 5.
modulo-2 ambiguity, i.e. to determine the integer portion of .
 3. ISSUES these a number of conclusions can be drawn. Table 2 compares
During the past 5 years a number of workshops have been held a number of characteristics of the two data types.
to discuss the developments in LiDAR and IfSAR and from

LiDAR IfSAR
Operational Acquired from fixed wing and helicopter platforms, Acquired from aircraft at high altitudes using single
at altitudes from 50m to 3500m. pass systems and from satellite platforms using
Space borne systems also in use. single and repeat pass systems.
Operates in day and night and in moist Operates in all conditions, although atmosphere can
atmospheres. affect accuracy.

Image Intensity image available with some sensors. Amplitude image created as part of system from
Frequently flown with digital camera. SAR.
Calibration Not fully developed. Well developed and essential.
Processing XYZ co-ordinates generated directly to form DSM. Complex processing now using mature algorithms
for DSM and orthoimage generation.
Layover and shadow will cause problems from
once only acquisition, but can be overcome by
multiple acquisition.
Coherence not a problem for single pass systems.
Post processing Produces DSM which needs processing to DTM. Produces DSM which need processing to DTM.
Good algorithms exist, but still not fully reliable. Significant editing still required.
Processing packages available. Response from different types of surface cover not
fully understood.
Characteristics of Density varies with sensor and altitude. Footprint larger than LiDAR, tends to smooth out
DSM Footprint size also varies. features.
Ground and tree surface can be seen with multiple Some penetration from trees, varies with frequency.
returns.
Gaps in data due to occlusions.
Accessibility to data Point cloud easily understood. Complex processing generally done by system
Many companies offer data acquisition and operator. SAR data not familiar to many people
processing. and nature of the DEM not always understood.
Software available for filtering, feature extraction Few operators.
etc.
Accuracy Best accuracy around 10cm in Z Best accuracy around 0.5m in Z
Applications Applications over limited areas with high accuracy. Suited to larger areas with lower accuracy.
2
Cost $500 per m $5 per m2
Standards LAS standard developed in USA. Specification Reliant on system operator.
and standards also being developed.

Table 2. Comparison of airborne LiDAR and IfSAR.

errors, of varying deflection errors, of time delays between

measurements, etc. It seems that it is much more essential to
Calibration of LiDAR data by the user is not well developed. understand the composition of a sensor system and what the
Normally a calibration surface will be measured prior to a flight manufacturer has done to avoid most of the effects.”
to determine bore sight alignment and any systematic error. A Katzenbeisser also suggests that general software for processing
datum shift can be applied to the measurements if necessary. real raw data (i.e. position, orientation and distance) may never
Some operators do not consider this to be necessary. In fact exist, as it would have to take into account a large number of
there are many sources of error which are not accounted for. parameters assigned with the individual manufacturing of a
Katzenbeisser (2003) has studied these and shows that “most of sensor system and which can not be generalized. Filin (2003)
the corrections, which might be applied have to be used at a has also investigated correction of systematic errors in LiDAR
very early stage of the data processing. Even if so called “raw and indicates the need for an error model for LiDAR.
data” (i.e. all echo coordinates) are available, the correction is
limited to GPS (or positioning) errors.” He goes to to write that The processing of the IfSAR data is now well established but
“The usual calibration flights (at the beginning and at the end of there are still problems, especially when high accuracy is
a survey) over flat terrain do not allow the detection of distance required. The main inherent problems are the presence of
layover and shadow, loss of coherence and the size of the
footprint.

The DSM is an accurate product derived directly from the

observations, however for many purposes a terrain model
(DTM), or bare earth model, is required. Much effort has been
expended to develop algorithms for this purpose, mainly
concentrating on LiDAR data. Sithole and Vosselman, (2003)
have reported on an ISPRS test of such filters. Less work has
been done on filtering IfSAR, where the scale is generally
smaller and the problems greater because of the footprint size
and the amount of penetration, or lack of it, of the microwaves
through vegetation.

Filtering algorithms generally incorporate a thresholding Figure 4. Multiple returns from a forest canopy.
function to decide whether a point lies on the terrain or on the © www.infoterra-global.com
observed surface. The threshold may depend on elevation of a
point or group of points, or it may depend on slope between Less work has been done on testing filtering of IfSAR DSMs.
adjacent points and these algorithms suffer from the problem of An evaluation of the Nextmap UK data was carried out at UCL.
assigning a value to the threshold. Figure 3 gives an example of This is discussed in detail in section 6. Figure 5 shows a
filtering from LiDAR carried out with the recursive terrain comparison of the Nextmap DSM, DTM and a GPS profile over
fragmentation filter (RTF) developed at UCL, (Sohn and an unvegeatated flood plane to the left and a a forest to the right.
Dowman, 2002). It can be seen that the forest has not been removed by the filter
used. Zhang et al (2004) have recently published an algorithm
developed specifically for IfSAR.

We can conclude that bare earth filtering still has problems and
that there will inevitably be a need for manual editing after the
automatic processing. Filtering of LiDAR is probably more
effective that that of IfSAR.

River Profile 3 - PT II Data,

Comparison GPS, DSM, DTM
a. Aerial Image. b. LiDAR DSM.
45
Elevation in meters

40
35
30 DTM
25 DSM
20
15 GPS
10
0 5 10 15 20 25 30 35 40 45 50 55 60 65
c. LiDAR DTM Distance along profile in meters

Figure 3. Filtering of LiDAR data using the RTF filter.

It can be seen that although the major surface features have Figure 5. Profiles across a flood plane and forest from the
been removed, the terrain is still not smooth. This is in part NextmapUK data.
due to small man made features, such as vehicles, and small
natural features, such as bushes, which fall below the assigned 6. ACCURACY
threshold. Sithole and Vosselman, (2003) found problems with
complex objects, attached objects, vegetation on slopes and The accuracy of both LiDAR and IfSAR is now quite well
discontinuities. Different filters cope differently with these established in empirical terms, but there are still error sources
problems. Smoothing filters can be used, but they can introduce which are not well understood or quantified, as discussed in
their own errors. section 4.

With LiDAR, some of these problems can be overcome if multi Ahokas et al (2003) have carried out an analysis of fixed wing
return systems are used. Figure 4 shows LiDAR returns over and helicopter LiDAR from different altitudes, over different
forest area, taken with an Optech 2033, in which the ground surface material and also looked at the effect of observation
surface can be confidently predicted from the last pulse return. angle. They concluded that ‘The analysis of the factors affecting
As point density increases, this becomes more reliable.
the total accuracy of the laser scanning is not as simple and as
straightforward as it was thought. ….. It was observed that there Photogrammetric check points collected from the stereo-model
is a flight line-dependent systematic and random error affecting of aerial photography in open bare earth areas, clear of
on the total accuracy obtained. It was observed that the higher surrounding surface features within a 5m radius were compared
the flight altitude, the higher is the random error of terrain with the DSM and the results can be found in Table 3. The
models. 800m flying altitude gives poorer results than 100m Nextmap DTM and the photogrammetric checkpoints are in
flying altitude. Laser measured heights are in general above the good agreement. A mean difference in elevation of -0.61m from
real ground surface. For asphalt surfaces a standard deviation of the check points and a rmse of 0.83m was observed.
10cm is obtainable from H=550m and from lower altitudes the
results are even better. A systematic error of typically 10 cm Furthermore, the vertical accuracy of the Nextmap data was
was observed due to observation angle changes.’ evaluated by comparing the Nextmap DTM bald earth surface
with Lidar derived reference DTMs. Results of these
Abdullatif et al (2003) have also investigated the accuracy of comparisons are also listed in Table 3. The Nextmap DTM was
LiDAR and report systematic errors, but an overall accuracy of subtracted from the reference DTMs (reference DTM minus
about 12cm. Overall accuracy of LiDAR can be as good as Nextmap DTM.
10cm, but in practice varies according to the quality of the
calibration and the terrain surface. Two sub areas of open terrain type were selected and difference
statistics produced. The Lidar DSM and the aerial photography
Airborne IfSAR can achieve accuracies of 0.5m, but also varies DSM were chosen as a reference. Both, the Nextmap DSM and
according to calibration, altitude and terrain surface. Mercer the DTM product were compared to the reference data sets. The
(2003a) discusses the trade offs between accuracy and swath results of these different comparisons are given in tables 4 and 5.
width and states that the theoretical accuracy from 30,000ft is
0.45m and 0.30 m at 10,000ft. The best accuracy of the Nextmap data is obtained over an open
field, which is interpreted as bare earth, where a mean
UCL has carried out an analysis of the Nextmap Great Britain difference between the Nextmap and aerial photography is
dat which used two test areas (Dowman et al, 2003) and made 0.23m (Nextmap higher) and the rmse is 0.43m. The mean
use of LiDAR, GPS and aerial photography as reference data. difference between the Aerial DSM and the Nextmap is
The initial comparison of Kinematic GPS with the Nextmap effectively zero. This suggests that the bare earth algorithm has
DSM showed unexpectedly large errors, which turned out to be removed a mean difference of 0.23m in bare earth area. This
due to the effects of hedges and trees on the Nextmap due to corresponds to the finding discussed earlier, which also
the footprint size. These were removed by filtering in order to indicates that the bare earth algorithm affects the mean. This
eliminate outliers due to vegetation that bias the accuracy result needs further investigation.
measures. The 3σ threshold was used as starting point for
filtering the difference data (KGPS minus Nextmap DSM). It is The Nextmap and Lidar surfaces are in good agreement in both
clear that points on the DSM are measured to be higher than the sub areas. Over a cropped area the Nextmap DSM has a
their true value because of the size of the footprint of the mean difference of –0.61m and rmse of 0.77m, from the Lidar
Nextmap data. If the bare earth algorithm is effective, these DSM. The Nextmap DTM has a mean difference of -0.38m and
errors should be corrected in the DTM, results are shown in rmse of 0.48m from the Lidar reference DTM.
table 3. It can be seen that a shift of between 0.3m and 0.8m has
occurred and that this has therefore significantly improved the .
root mean square error.

Comparison Terrain Type Land cover n vmin vmax vMean σ[m] Rmsez [m]
KGPS 3 vs. Nextmap
Mixed (hilly, KGPS located along
DSM points > ±1.5m 1994 -1.50 0.05 -0.95 0.34 1.00
flat) roads (bare earth)
removed
KGPS located along
KGPS 3 DTM vs. Mixed (hilly,
road network (bare 2647 -1.52 0.48 -0.66 0.32 0.73
Nextmap DTM flat)
earth)
KGPS6 DSM vs. KGPS located along
Mixed (hilly,
Nextmap DSM points road network (bare 1475 -1.85 1.00 -0.96 0.49 1.08
flat)
> ±1.9m removed earth)
KGPS located along
KGPS 6 DTM vs. Mixed (hilly,
road network (bare 1568 -1.73 3.58 -0.14 0.45 0.47
Nextmap DTM flat)
earth)
Air photo check points
Mixed Bald earth 66 -1.66 0.43 -0.61 0.57 0.83
vs. Nextmap DSM
Lidar DTM vs. Mixed (hilly,
Bald earth 85362 -9.20 12.04 -0.22 0.10 1.01
Nextmap DTM (5) flat)

Table 3. Summary of results from NUI Nextmap DTM evaluation of Shrewsbury area
Notes: All DEMs have 5m grid. KGPSi refers to ith profile recorded along roads.
 Sub Terrain Land Rmsez
Comparison n vmin vmax vMean σ[m]
Area Type cover [m]
Lidar DTM (5m) vs. Nextmap hilly
1 crops 11484 -1.55 0.52 -0.38 0.30 0.48
DTM (46-61m)

Lidar DTM (5m) vs. Nextmap Flat

2 crops 2166 -1.14 0.04 -0.41 0.21 0.46
DTM (49-53m)

Table 4. Lidar compared to Nextmap DTM (selected areas)

Terrain Land Rmsez

Sub Area Comparison n vmin vmax vMean σ[m]
Type cover [m]
Aerial DSM vs. Nextmap hilly
1 crops 11484 -10.01 1.094 -0.31 0.44 0.54
DSM (46-61m)
Aerial DSM vs. Nextmap hilly
1 crops 11484 -0.97 0.54 -0.07 0.20 0.21
DTM5 (46-61m)

Aerial DSM vs. Nextmap Flat

2 crops 2166 -1.38 0.99 -0.23 0.37 0.43
DSM (49-53m)
Aerial DSM vs. Nextmap Flat
2 crops 2166 -0.54 0.37 -0.00 0.17 0.17
DTM (49-53m)

Table 5. Aerial photography compared to Nextmap DTM (selected areas)

0.22m difference with a rmse of ±1.01m, from the Lidar

The areas chosen for validation in Shrewsbury and Worcester DTM.
cover a flood plain of the River Severn with some relief of
about 60m above the river. It also contains a variety of land • The GPS profiles along the roads show good agreement
cover types, including bare ground, crops, woodland and built with the Nextmap DTM data when the effects of trees and
up suburban areas. hedges have been removed.

The main conclusions are as follows: The evaluation of the Nextmap Great Britain data reveals a
number of characteristics of the IfSAR and LiDAR data, and of
• The vertical accuracy of the Nextmap data varies according the filtering techniques, which are general to this type of data.
to the type of the terrain where the comparison is made, It also reveals unexplained differences which need further study.
and in particular the land cover. For example, it is known
that forest and dense urban areas significantly decrease
vertical accuracies of Digital Elevation Models. In general
it can be stated that the mean surface of the Nextmap data The validation of any DEM is clearly very important. This
is higher than the reference data. This is expected because however can be quite difficult and expensive when the precision
of the size of the Nextmap footprint, and the general effect of the product is so high, and interpolation is necessary in the
of vegetation in IfSAR measurements. The elevation process. Some operators have high confidence in their product
measured for any IfSAR DSM sample (square footprint and do not consider validation necessary. The quality of the
somewhat larger than the 5m DSM sample distance) result positioning can be checked from the GPS record. The normal
from a combined signal of scattering objects. Thus, raised procedures for validation include the use of reference data such
objects such as trees and hedges located in the sample area, as check points located as targets or points on open surfaces,
contribute to the elevation value measured. reference DEMs or profiles. Checks can also be made for
consistency and for outliers, and correlations can be
• When comparing the Nextmap DSM with investigated between the data and vegetation or slope. Where
photogrammetric checkpoints, which were measured in the point density is high enough, targets provide an very good
open terrain, a mean difference in elevation of -0.61m and validation surface for LiDAR. The Highways Agency (HA) in
an rmse of ±0.83m were observed. the UK specifies that boards, 1.2m x 1.2m, centred over a co-
ordinated point and accurately levelled must be used. The HA
• The best accuracy of the Nextmap DTM is obtained over requires a point density of 7 – 10 points per m2, and thus about
an open field, which is interpreted as bare earth, where a 10 points are expected to fall on the board, allowing significant
mean difference between the Nextmap and aerial statistics to be generated. An illustration is shown in figure 6.
photography was –0.001m and the rmse was ±0.17m
(Nextmap higher). For IfSAR this is not usually appropriate but traditional corner
reflectors can be used. Because of the footprint size of IfSAR
• The Nextmap and Lidar surfaces are in good agreement, comparisons over large areas or profiles are better suited.
overall. Over a cropped area the Nextmap DSM had a Kinematic GPS profiles along roads have proved to be very
mean difference of –0.61m and a rmse of ±0.77m, from the useful for checking both LiDAR and IfSAR. Examples are
Lidar DSM. Over the whole area the Nextmap DTM was - given in Morley et al (2000) for the Landmap project and
Dowman and Fischer (2003) for Nextmap UK.
 Spaceborne IfSAR is more established as a source of DEMs that
is airborne. The ESA ERS Tandem mission has acquired very
wide coverage of interferometric SAR pairs and this is much
used for the generation of regional DEMs. For example the
Radarmap of Germany produced by DLR (Kosmann et al, 1994)
and the Landmap project in UK (Morley et al, 2000). The
Shuttle Radar Topography Mission (SRTM) has also produced
DEMs and orthoimages between 60˚ North and 56˚ South. In
addition RadarSat, JERS, and ENVISAT all produce
interferometric data and in the future RadarSat 2 and ALOS
PALSAR will join the ranks of IfSAR data generators. IfSAR
has also had an important application in differential mode for
monitoring tectonic movement and subsidence.
Figure 6. Specification of the UK Highways Agency control
panels with location along the highway indicated. With the exception of SRTM, satellite IfSAR uses repeat pass
data, and this can suffer from the problem of poor coherence
and atmospheric effects, which degrade the data and can cause
gaps in the DEM. SRTM also suffers from problems,
particularly layover in mountainous areas, and the SRTM
There is a clear separation between the costs and the dataset does contain some gaps.
applications of airborne LiDAR and IfSAR. LiDAR is
generally high cost and high accuracy, and suited to covering The accuracy of IfSAR DEMs from spaceborne platforms varies
small areas. IfSAR is less accuracy but can cover large areas significantly, depending on the coherence, itself dependent on
very economically. Mercer (personal communication) quotes the interval between acquisition of the two images and stability
costs of about $5 per km2 for the IfSAR data of Nextmap of the weather and atmosphere, and the terrain.
Britain and about $500 per km2 for Airborne LiDAR. Wulder
(2003) quotes Canadian $1900 per km2 for LiDAR with 30cm
posting. Some of the shortcomings of IfSAR can be overcome
at additional cost. For example problems with occlusions can Data fusion exploits the synergy of two or more data sets to
be solved if an area is flown from two look directions. create a new data set which is greater that the sum of the parts.
The ISPRS Journal of Photogrammetry and Remote Sensing
There are some important points to consider when looking at the (Vol 58(1-2), 2003), published a theme issue on multi-source
accuracy of products: data fusion for urban areas which clearly demonstrates the range
and importance of data fusion. Data fusion can be used for
• The product must be fit for purpose with suitable many applications. Some of the established ones are:
accuracy and point density • Assisting phase unwrapping
• The data must fit to other data - very important with • Eliminating errors and blunders
web delivery to non specialist users • Atmospheric correction
• Accuracy must be specified • Providing orientation in areas where there is no
control
• Terrestrial images to LiDAR
• Feature extraction, such as buildings and roads
As LiDAR becomes more widely used, it becomes more
• Other aspects of feature extraction and environmental
essential that data can be easily read by standard mapping
analysis (see ISPRS Journal 58(1-2)).
systems, and passed between users. This implies a need for
standards and interoperability.
Some examples of how DEM data from LiDAR or IfSAR can
be combined with other imagery or map data for feature
In the USA a standard has been established for the formatting
extraction are given in section 13.5. Honikel (2002) shows how
of LiDAR data (http://www.lasformat.org/). The LAS format
ERS If SAR and SPOT DEMs can be fused and develops a
has been accepted by the American Society for Photogrammetry
theory for this; Csanyi and Toth (2003) also discuss the
and Remote Sensing (ASPRS) and is being widely used This
theoretical aspects of merging IfSAR and LiDAR data.
has also been incorporated into the Highways Agency
Specification for LiDAR in the UK. In the USA, FEMA
SRTM provides another interesting case study. Because SRTM
(2000) has a standard specification for LiDAR and ASPRS are
provides a near global data set which is geocoded with accuracy
developing a handbook of operational LiDAR Mapping. ISO
which is better than any other comparable global DEM, it can
19130 TC 211 "Geographic information / geomatics" project
be used to give initial orientation for higher accuracy data and
team of "Sensor and data models for imagery and gridded data"
be used to assist with phase unwrapping and atmospheric
has published a Committee Draft of an ISO standard for
correction.
photogrammetry and remote sensing but this does not include
sensor models for LiDAR and SAR.
!
It is important that these initiatives are carried forward; this is
an opportunity to define standards for a new sensor and new LiDAR and IfSAR are relatively new products and it is
products at an early stage in their development. therefore necessary to overcome the reluctance of users to make
use of them. Educated users, especially research organisations, 13.3 Environmental applications
will always be keen to look at new products, but users more
familiar with photographic products will take more time. Some A major application for environmental use is forestry. The
of the major users of both airborne LiDAR and IfSAR have ScandLaser Scientific Workshop of Airborne Laser Scanning of
been new users, for example power generation companies for Forest, held from September 3-4, 2003 in Umea, Sweden, gives
powerline survey, and insurance companies for assessing flood a very detailed view of the current status of LiDAR for forestry,
risk. But there can be problems with such users not e.g Wulder (2003), Naesset (2003), Hyyppa et al (2003).
understanding the characteristics of the data, nor the accuracy
which can be expected. Hamdan (personal communication) has noted that Dubayah and
Drake (2000) listed the key forest characteristics that can be
LiDAR can produce a high density of points, and although this measured directly or indirectly by LiDAR. Among the
is an advantage in some situations, it can also cause problems, parameters that can be retrieved directly are canopy and tree
for example in the volume of data to handle. High density height, timber volume, forest mixtures according to tree species,
might be necessary to identify detail on the ground, small natural age classes, forest canopy closure, decision of forest /
gullies or crash barriers on highways for example, but not on non-forest and sub canopy topography. Beside this, above
the main carriage way. Thus there is a problem on how to this ground biomass and volume, basal area, mean stem diameter,
the data to retain only what is needed. Intensity images may vertical foliar profiles, canopy volume and large tree density can
appear to be useful in order to make it unnecessary to fly a either be modelled or inferred from LiDAR measurements.
camera as well, but their quality is not as good, and there is no Other important parameters for forest such as canopy cover, leaf
standard for measuring intensity. On the other hand flying a area index (LAI) and life form diversity need different approach
camera with the LiDAR can be a disadvantage as it means that where data fusion from lidar and other sensor is essential. In this
the lighting conditions must be good enough for the camera, case, the vertical component provided by LiDAR should be
whilst the LiDAR could operate in poorer lighting conditions. fused with information from passive optical, hyperspectral,
thermal and radar remote sensing (Hill et al.,2003). Apart from
IfSAR is a complex system and users do not need know the that, LiDAR data like other optical remote sensing techniques
intricacies of the processing, but they do need to understand are restricted by clouds and dense atmospheric haze. This can
that SAR samples a footprint which is quite large and that attenuate the signal before it reaches the ground. Another
different types of land cover give different responses. They also limitation of LiDAR is the lack of algorithms and data
need to understand the meaning of orthocorrection, (terrain processing expertise required for operational use of the data. All
orthoimages and true orthoimages), the need for compatibility these enhance the integration of this data with other satellite
of projection and datum, and the significance of error statistics. system.
In other words the users need to be educated to some degree and
the data provider needs to ensure that they are. An interesting new development is the combination of airborne
LiDAR with terrestrial LiDAR for forestry and the creation of
In the United Kingdom, the Highways Agency has produced a virtual forest environments. (Evans, 2003). Off shore tidal area
specification for LiDAR surveys which has been produced in are anther important application area. A LiDAR survey has
close consultation between the data providers and the client. been done for Willapa Bay in Washington, USA, demonstrating
This ensures that the client gets what is needed, for example in the utility of the technique in intertidal areas.
terms of data formats, and visualisation of products to help new
users, and the provider understands what is required. 13.4 Engineering applications

" LiDAR has been used for engineering work such as railways,
powerlines and highways because of its high vertical accuracy
13.1 Introduction and the density of points. The application for power lines,
(Silver, 2001) and the ability to accurately determine the
There are now a great many applications for DEMs from position of the cables is an excellent indication of the usefulness
LiDAR and IfSAR data and it is beyond the scope of this paper of LiDAR.
to deal with all of them. We will therefore briefly review some
of the innovative applications and concentrate on those which When a camera is flown with the LiDAR, even if a non metric
involve the use of data from more than one source. camera, then large scale mapping can be carried out. Figure 7
shows a plot of a highway intersection with detail and contours.
13.2 Regional and global mapping Compiled from the LiDAR DEM and a digital image acquired
at the same time as the LiDAR data.
IfSAR has proven itself for low cost DEM generation over large
areas. The prime example is SRTM, but large areas have also The use of LiDAR for the generation of 3D city models is well
been mapped with ERS data, for example the Radarmap of established and some techniques are discussed in section 13.5.
Germany (Kosmann et al, 1994). Airborne systems have been High density point clouds can be used to extract buildings and
used for generation of DEMs and orthoimages over large areas roof detail by fitting planes to the points. TerraScan provides
such as the Nextmap Britain project (Mercer, 2003a, Dowman tools for creating fully dimensional vectorised models of
and Fischer, 2003). The Nextmap data was originally buildings from LiDAR data based on identification of planar
commissioned for an insurance company for flood risk analysis, roof surfaces. Chayakula (2004) has investigated the use of
but is now being used more widely than that, and is airborne IfSAR in urban areas and shown that useful
complementary to LiDAR, which is useful in denser urban information can be extracted. Houshmand and Gamba (2001)
areas. Intermap have carried out IfSAR surveys in many parts have also worked on this topic (see below).
of the world including Malaysia and Indonesia, and are starting
on a coverage of the whole of the USA.
 It can be seen that LiDAR and IfSAR are important new sources
of data for generating geospatial products. They have opened
up new markets by filling gapes which could not be filled by
aerial photography or optical satellite imagery. It can also be
seen that LiDAR and IfSAR are themselves complementary,
and also complementary to other sources of data. It has been
shown that IfSAR is more economical for wide area coverage,
provides an intensity image, which can be orthorectified, a
coherence image and multifrequency and multipolarised data
which can give more information about the land cover than
hitherto possible. LiDAR on the other hand gives a high
density cloud of 3D points which can accurately define both
elevation and plan position. There are however a number of
restriction on wider use of LiDAR and IfSAR and open
questions on their future development.
Figure 7. Detail plot from LiDAR and digital photo.
Although LiDAR has the potential for application in building
13.5 Combining data for feature extraction
extraction and 3D city modelling, automatic feature extraction
is still not mature and therefore the output is unreliable, and
LiDAR is very important for feature extraction and has been
manual editing is very expensive. LiDAR is also very
widely used with other data sources for this purpose. Haala &
expensive for small areas. Wider use of LiDAR may therefore
Brenner (1999a) have demonstrated the combination of LIDAR
have to wait until better feature extraction algorithms are
data with multispectral aerial images for the automatic
available. New airborne technologies such as 3 line optical
classification of buildings and trees. Haala and Brenner (1999b)
sensors could also compete with LiDAR when they become
have also shown automatic 3D building reconstruction in a
more mature and can acquire data with higher resolution than at
system which combines 2D GIS data and LiDAR. Based on
present. Three line data avoids occlusions and adds redundancy
given outlines of the respective buildings which were integrated
to the data set. Multi sensor data could also do this. The use of
with dense surface data from airborne LIDAR measurements,
a digital camera with LiDAR is already commonplace, but a
virtual city models were created for an extensive number of test
good model for reconstruction and error analysis is needed. In
sites using this software. Sohn and Dowman (2004) have
order to inspire confidence in the data better theoretical models
combined low density LiDAR with high resolution satellite
are required, both for single sensors, and for data fusion, in
sensors to extract buildings. Dell’Acqua et al, (2003) have
order that the errors can be better understood. Potential errors
combined LiDAR with IfSAR in urban areas. The paper shows
such as multipath and transparency effects also need to be
that it is possible to exploit LIDAR DEM to improve to some
studied much more. More comparative tests, especially with
extent the two- and three-dimensional representation of
different algorithms, need to be carried out, although this is now
buildings extracted from IfSAR data. The method helps in
happening through ISPRS (Vosselman and Sithole, 2003) and
recovering building displacement and distortion due to the side-
EuroSDR (http://www.oeepe.org/2002/index.htm), for example.
looking nature of radar. This is shown in figure 8.
IfSAR could also benefit from comparative testing and the
establishment of international test sites would be beneficial.
CEOS and EuroSDR could contribute to this.

A better understanding of the quality measures and error

statistics, and development of understandable uncertainty
measures would also be beneficial. Organisation such as USGS
and Ordnance Survey are establishing good practice in this area
by showing heritage and uncertainty in their data; this should be
encouraged, and their methods publicised.

15. NEW DEVELOPMENTS

A number of clear trends have been identified. These include

more accurate, higher density data, new applications and the
development of new algorithms. High density LiDAR point
clouds mean that buildings can be better extracted, and details
of highway surfaces and infrastructures can be surveyed more
easily.

The main developments in progress are in the area of IfSAR,

particularly the development of multi frequency, multi polarised
SAR. The ESA spaceborne ENVISAT mission is already
Figure 8. 3D characteristic of buildings from IfDSAR and operating the ASAR sensor and other mission such as the
LiDAR. After Dell’Acqua et al, (2003). Japanese ALOS PALSAR will be launched shortly. Satellite
SAR resolution is being improved, as is the positional and
orientation accuracy. Constellations of small satellites are to develop threoretical models for sensors and data fusions, to
planned which will extend the availability and flexibility of improve bare earth filtering and to improve feature extraction.
IfSAR data. User need to be educated more and to aid the greater use of the
data, standards need to be defined for products and for data
GeoSAR is a dual-frequency, dual-polarimetric, interferometric exchange.
airborne radar mapping system that generates DEMs and REFERENCES
orthorectified radar reflectance maps near the tops of trees as
well as beneath foliage, Abfdullatif A, Bethel J, Mikhail E, (2003). Accuracy and
(http://southport.jpl.nasa.gov/html/projects/geosar/geosar.html). assessment of airborne laser scanning data. Proceedings of
The GeoSAR system collects radar data in two frequencies. The ISPRS Joint Workshop of ISPRS WG I/3 and II/2, Three
X-band maps the first surface, near the top of trees and the P- dimensional Mapping from InSAR and LIDAR, Portland
band maps beneath the foliage and assists in the production of a Oregon, USA 17th – 19th June 2003. CDROM.
bare-earth terrain model and the detection of structures beneath
trees. Mercer (2003b) reports on the use of Polarimetric P band Ahokas E, Kaartinen H, Hyyppä J,.(2003). A Quality
data for generation of DEMs of forest areas, and compares this Assessment of Airborne Laser Scanner Data. International
with X band InSAR. Archives of the Photogrammetry, Remote Sensing and Spatial
Information Sciences, XXXIV(3/W13) CDROM
We are beginning to see the economy of airborne IfSAR being
used for wide area DEMs to complement the global data from Baltsavias E P (1999a), A comparison between photogrammetry
SRTM. These can be created from a single source, hence and laser scanning, ISPRS Journal of Photogrammetry &
providing a homogeneous data set, generated over a short period Remote Sensing 54:83-94
of time. The use of permanent scatterers to monitor subsidence
with IfSAR is also being developed (Ferretti et al, 1999) Baltsavias E P (1999b), Airborne laser scanning: basic relations
and formulas; ISPRS Journal of photogrammetry & Remote
A better control infrastructure is becoming established to allow sensing 54:199-214
woder and easier use of GPS, INS systems. National mapping
agencies are establishing permanent continuous recording GPS Chayakula T, (2004). Airborne SAR/IfSAR for Mapping in
stations, but operators prefer to set up their own base stations. Urban Areas. Ph D thesis, University of London.
The introduction of Galileo will further extend the use of
positioning systems. Csanyi N, Toth C, (2003)). Combining LiDAR Data with
Stereoscopically Extracted Surfaces: Feature Level Fusion
The importance of validation and improved quality assurance is Proceedings of ISPRS Joint Workshop of ISPRS WG I/3 and
being recognised and the introduction of internationally II/2, Three dimensional Mapping from InSAR and LIDAR,
accepted standards is being discussed. Portland Oregon, USA 17th – 19th June 2003. CDROM.

As the data becomes more widely used, new image processing Dell’Acqua F, Gamba P, Lisini G (2003). 3D urban mapping
systems are becoming available. TerraSolid is widely used now jointly exploiting InSAR and LiDAR DEMs. Proceedings of
for processing LiDAR data, and more tools are becoming ISPRS Joint Workshop of ISPRS WG I/3 and II/2, Three
available. Packages such as eCognition are particularly suited dimensional Mapping from InSAR and LIDAR, Portland
to use with SAR data and DEMs. Intelligent systems such a Oregon, USA 17th – 19th June 2003. CDROM.
ALFIE, (Automated Linear Feature Information Extraction), a
new system for generating simulations for military use being Dowman,I, Balan P, Renner K, Fischer P, (2003). An
developed in the UK (Wallace et al, 2004). The system is based Evaluation of Nextmap Terrain Data in the Context of UK
on existing algorithms integrated into a toolkit within a National Datasets. Report to Getmapping 19 pages.
processing environment which can automatically select which
tools to use with particular data for specified applications, and Dowman I, Fisher P, (2003). Evaluation of IfSAR and Lidar
which can also make use of context in extracting features. Data for Flood Risk Assessment. Proceedings of ISPRS Joint
ALFIE is also linked to an object oriented data base and works Workshop of ISPRS WG I/3 and II/2, Three dimensional
with a developed feature extraction environment. Mapping from InSAR and LIDAR, Portland Oregon, USA 17th
– 19th June 2003. CDROM.
16. CONCLUSIONS
Dubayah R.O. and Drake J.B. (2000) LiDAR remote sensing for
It has been shown in this paper that LiDAR and IfSAR are now forestry, Journal of Forestry 98:44–46
widely used and that this type of data is opening up new
markets and new opportunities in areas such as powerline Evans D, (2003). LiDAR Assessments of Forest Characteristics:
surveys, flood risk mapping and large area mapping. The two Scaling up to Landscape-level Forest Mapping and Monitoring
types of data are complementary with each other and each can Proceedings of ISPRS Joint Workshop of ISPRS WG I/3 and
be used with other data sources to generate new value added II/2, Three dimensional Mapping from InSAR and LIDAR,
products. Portland Oregon, USA 17th – 19th June 2003. CDROM.

That having been said, we have also shown that there are still FEMA (2000), Flood Hazard Mapping, Airborne Light
problems with using the data and more development needed Detection and Ranging systems, Appendix 4B, Federal
before the technology is fully mature. The calibration of Emergency Management Agency, 2000, FEMA 37,
LiDAR data is not well developed, and neither are http://www.fema.gov/mit/tsd/lidar_4b.htm
specifications or quality assurance techniques. The generation
of bare earth models (DTMs) are still liable to error and manual
editing is still needed. The main areas for further research are
 Ferretti A, Rocca F, Prati C, (1999). Permanent Scatterers in II/2, Three dimensional Mapping from InSAR and LIDAR,
SAR Interferometry. Proceedings of IGARSS99, Hamburg Portland Oregon, USA 17th – 19th June 2003. CDROM.
Germany, June 28th – July 2nd 1999.
Mercer, B J (2002), Comments on GCP Selection in Support of
Filin S (2003). Recovery of systematic biases in laser altimetry STAR-3i DTM Validation, Intermap in house publications,
data using natural surfaces. Photogrammetric Engineering and April 17, 2002, www.intermap.ca
Remote Sensing, 69(11):1235-1242.
Mercer B J, Schnick S. (1999), Comparison of DEMs from
Haala, N. & Brenner, C. (1999a). Extraction of buildings and STAR-3i Interferometric SAR and Scanning Laser,
trees in urban environments. ISPRS Journal of Photogrammetry International Archives of Photogrammetry and Remote Sensing,
& Remote Sensing 54(2-3):130-137 Vol.32, Part 3W14, La Jolla, Ca, 9-11 November 1999
Haala, N. & Brenner, C. (1999b). Virtual City Models from
Laser Altimeter and 2D Map Data. Photogrammetric Morley, J.G., Muller, J.-P.A.L., Walker, A.H., Kitmitto, K.,
Engineering & Remote Sensing 65(7):787-795. Mitchell, K.L., Chugani, K., Smith, A., Barnes J., Cross, P.A.
And Dowman, I.J., (2000). LANDMAP: Creating a DEM of the
British Isles by SAR interferometry., International Archives of
Hill R.A, Hinsley S.A, Bellamy P.E, Balzter H. (2003)
Photogrammetry and Remote Sensing, 33(B4/2):686-693.
Ecological Applications of Airborne Laser Scanner Data:
Woodland Bird Habitat Modelling, Proceedings of the
Naesset E, (2003). Laser scanning of forest resources – the
ScanLaser Scientific Workshop on Airborne Laser Scanning of
Norwegian experience. Proceedings of the ScandLaser
Forests, pp. 78 - 85.
Scientific Workshop of Airborne laser Scanning of Forest,
Septemeber 3-4, 2003 Umea, Sweden, Working Paper 112 2003
Honikel M, (1998) Fusion of Optical and Radar Digital
Swedish University of Agricultural Sciences, p p34-41.
Elevation Models in the Spatial Frequency Domain. 2nd Int.
Workshop "Retrieval of Bio- and Geo-physical Parameters from
Silver A, (2001). Transmission line verification and and
SAR data for Land applications", 21-23 October 1998, ESTEC,
rerating services. Proceedings of ISPRS WGII/2, Three
Noordwijk, The Netherlands pp.537-543.
Dimensional mapping from InSAR and LIDAR, Banff Canada
11-13th July 2001. CDROM.
Houshmand B, Gamba P, (2001). Interpetation of InSAR
mapping for geometrical structures. Prod of IEEE/ISPRS Joint
Sohn G, Dowman I J (2002), Terrain surface reconstruction by
Workshop on Remote Sensing and Data Fusion over Urban
the use of tetrahedron model with the MDL criterion.
Areas, Rome, 8-9 Nov 2001, pp 309-311.
International Archives of the Photogrammetry, Remote Sensing
and Spatial Information Sciences, 34(3A):336-344.
Hyyppa J, Hyyppa H, Maltamao M, Yu X, Ahokas E, Pyysalo
U, (2003). Laser scanning of forest resources - some of the
Sithole G and Vosselman G, (2003) Comparison of Filtering
Finnish experience. Proceedings of the ScandLaser Scientific
Algorithms. International Archives of the Photogrammetry,
Workshop of Airborne laser Scanning of Forest, Septemeber 3-
Remote Sensing and Spatial Information Sciences,
4, 2003 Umea, Sweden, Working Paper 112 2003 Swedish
XXXIV(3/W13) CDROM
University of Agricultural Sciences, pp 52-58.
Wallace S, Hatcher M, Ley R, Dowman I, Sohn G, (2004).
Katzenbeisser R, (2003). About the Calibration of LiDAR Automatic Extraction and Classification of 2D and 3D Features
Sensors. International Archives of the Photogrammetry, from High Resolution Satellite Imagery. International Archives
Remote Sensing and Spatial Information Sciences, of Photogrammetry and Remote Sensing, 35 (in Press)
XXXIV(3/W13) CDROM
Wulder M, (2003). The current status of laser scanning of
Kosmann, D., Winter, R., Sties, M., Wiggenhagen, M., 1994 forests in Canada and Australia. Proceedings of the ScandLaser
Mosaicing and Classification for the Radarmap Germany. Scientific Workshop of Airborne laser Scanning of Forest,
Proceed. Second ERS-1 Symposium - Space at the Service of Septemeber 3-4, 2003 Umea, Sweden, Working Paper 112 2003
our Environment, Hamburg, 11-14.10.1993, ESA SP-361, Swedish University of Agricultural Sciences, p p20-32.
Noordwijk, pp. 629-633.
Zhang Y, Tao V, Mercer B, 2004. An initial study on automatic
Mercer B, (2003a). Proceedings of ISPRS Joint Workshop of reconstruction of grounds DEMs from airborne IfSAR DSMs.
ISPRS WG I/3 and II/2, Three dimensional Mapping from Photogrammetric Engineering and Remote Sensing, 70(4):427-
InSAR and LIDAR, Portland Oregon, USA 17th – 19th June 438.
2003. CDROM.

Mercer B, (2003b). Extraction of ground DEMs beneath forest

canopy using P-band polarimetric InSAR PolInSAR.
Proceedings of ISPRS Joint Workshop of ISPRS WG I/3 and

View publication stats