IKONOS Basemap White Paper - Geographic Resources Center

1
Planimetric accuracy of Ikonos 1-m panchromatic image products
and their utility for local government GIS basemap applications
C.H. DAVIS and X. WANG
Department of Electrical Engineering
University of Missouri - Columbia
Columbia, MO 65211 USA
Phone: (573) 884-3789
FAX: (573) 882-0397
DavisCH@missouri.edu
XiangyunW@hotmail.com
Abstract. A detailed assessment of the planimetric accuracy of high-resolution (1 m)
panchromatic Ikonos orthoimage products for three different test sites located in the State of
Missouri is presented. The main objective of this study was to evaluate the potential of Ikonos
orthoimage products for use as digital image basemaps in local government GIS systems. For
maximum utility and adoption in a wide variety of local government planning and management
applications, a planimetric accuracy of 3-4 m CE90 is considered nominal. Evaluation of the
planimetric accuracies of Ikonos orthoimages acquired under a Space Imaging/NASA Databuy
agreement were found to be 1.1 m and 1.7 m CE90 at two different test sites. The terms of the
NASA Databuy agreement require a 2 m CE90 planimetric accuracy. Thus, our independent
assessment confirms that the Space Imaging products met this NASA Databuy specification.
Low precision, low cost georeferenced Ikonos image products (Carterra Geo, 50 m CE90)
were orthorectified using third party commercially available software and various custom NAPP
DEMs. The planimetric accuracies of the resulting Ikonos orthoimages were found to vary
between 2-4 m CE90. In addition, USGS DEMs were also used to orthorectify georeferenced
Ikonos image products. The resulting Ikonos orthoimages were found to have planimetric
accuracies from 2-7 m CE90. Planimetric accuracies of 2-3 m CE90 can be obtained from
georeferenced Ikonos using USGS DEMs with RMS vertical accuracies on the order of 2 m.
The approach demonstrated here can be used to deliver up-to-date, cost effective orthoimages
from Ikonos Carterra Geo products that yield planimetric accuracies suitable for use as digital
image basemaps by local governments.
1. Introduction
Urban growth and change places a heavy demand on local governments to seek better
planning and management approaches. Increasing urbanization puts pressure on natural
resources and existing infrastructure. Decision makers within various local government bodies
must deal with a wide-variety of issues that routinely have economic, social, and political
consequences. In addition, state and federal governments have issued an increasing number of
regulations that mandate the monitoring and tracking of numerous issues by local governments.
Elected officials in these local governments require timely information products to support
policy decisions on issues that are often interrelated and can span several political boundaries.
Growth assessment, infrastructure inventory and planning, environmental assessment, and risk
management impact and drive policy decisions for these managers.
2
As a result, local governments have invested considerable resources in developing
Geographic Information Systems (GIS) to aid them in their planning and decision making
processes. A digital image basemap is a key information layer in many local government GIS
systems. Image basemaps are used by city planners and engineers for tax assessment, inventory,
construction planning (roads, bridges, etc.), stormwater management, and other civil planning
activities (greenbelt preservation, emergency 911, etc.).
A major stumbling block to the effective application of remote sensing imagery for digital
basemap generation is the positional accuracy of the imagery. Existing vector data layers (road
centerlines, parcel and zoning boundaries, etc.) are routinely superimposed upon the image
basemap for planning and assessment applications. Figure 1 shows a typical situation where an
existing GIS data layer (parcel boundaries) is overlain on an imagery basemap that has a lower
positional accuracy. If vector data layers do not line up with the image basemap, then the
basemap is perceived to be of limited value and will not be integrated into standard operations
and decision-making processes within the local government. Thus, the horizontal or planimetric
accuracy of a remote sensing image is a critical measure of its utility for application as a digital
basemap in local government GIS systems.
2. Digital basemap sources
Orthorectified digital images derived from aerial photographs are often used as basemaps in
many local government GIS systems. Image orthorectification using an appropriate Digital
Elevation Model (DEM) to correct for horizontal errors (due to terrain displacement) is required
to produce a basemap with a high degree of positional accuracy. The most widely available
high-resolution digital orthoimage dataset available to local governments in the USA are the
Digital Orthorectified Quarter Quadrangles (DOQQs) produced by the United States Geological
Survey (USGS 1996). The DOQQs are derived from 1:40,000-scale National Aerial
Photography Program (NAPP) panchromatic aerial photographs acquired about every 5-6 years
(Light 1993).
The NAPP photographs are acquired with 60% N/S overlap to enable stereo processing for
DEM extraction. The USGS 7.5-minute Level 1 DEMs are derived from stereo-processing of
the NAPP imagery. The USGS DEMs have a 30-m horizontal resolution, a nominal RMS
vertical accuracy of 3-5 m, and a worst-case vertical accuracy of 7 to 15 m (USGS 1997). The
USGS DEMs have been used to orthorectify the original NAPP photographs to produce DOQQs.
The DOQQs have a ground sample distance or pixel size of 1 m and a worst-case planar
accuracy of 10 m CE90. The reported 10-m planar accuracy is a circular error at 90%
probability (CE90) and corresponds to US National Map Accuracy Standards (NMAS). NMAS
specifies that 90% of well-defined points must fall within the specified planar accuracy of the
image or map.
The DOQQ pixel size is sufficient for local government basemap applications. The worstcase planar accuracy is a limitation for widespread use. For example, typical road widths are on
the order of 10 m, so a 5 m planar accuracy (CE90) would be required for road centerline vectors
to lie within the road width. Nevertheless, many city and county governments have utilized
3
DOQQs as their data source for orthoimage basemaps because the cost to obtain these data is
essentially zero. However, problems with positional displacements between individual DOQQs
in the overall basemap and problems with proper registration of vector data layers are routinely
experienced. Another limitation of the DOQQs is that these are produced from NAPP aerial
photographs that are only acquired about every 5-6 years. Thus, the DOQQs are often outdated
in many areas and do not reflect the current situation within the city and/or county, especially in
areas that experience even moderate urban development and/or change.
The recent launch of the Ikonos satellite has opened up a new area for acquiring up-to-date
high-resolution panchromatic digital imagery for use as digital orthoimage basemaps. Ikonos
panchromatic imagery has a nominal pixel size of 1 m and an 11-bit information content. The 1m pixel size is identical to the DOQQ pixel size while the 11-bit imagery provides image
contrast and quality that is superior to the DOQQs. The Carterra Geo Ikonos product costs
$21/km2, bit its horizontal precision is poor with a planar accuracy of only 50 m CE90 (Space
Imaging 2000). While this product is affordable for most local government entities, the planar
accuracy of 50 m CE90 renders this product unusable for traditional basemap applications.
Our interaction and discussion with basemap users within various local government user
communities indicates that planar accuracies of 3-5 m CE90 are required for 1-m resolution
digital basemaps to be useful for GIS applications. The Carterra Precision Ikonos panchromatic
image product costs $63/km2 and has a planar accuracy of 4 m CE90 (Space Imaging 2000).
While this meets the planar accuracy requirement for basemap implementation, it is three times
more expensive than the Carterra Geo product. This cost can be prohibitive for many local
government entities with limited fiscal resources. For example, a typical county within the State
of Missouri, USA covers an area of approximately 2,400 km2. Thus, for complete county
coverage the Geo and Precision Ikonos products would cost $50,400 and $151,200 (US),
respectively. The former figure is cost-competitive with traditional aerial photographic surveys,
while the latter figure is beyond the budgetary capability of most local county governments. The
challenge then becomes to develop a methodology that would enable the use of the lower-cost
Carterra Geo products for creation of digital basemaps.
In this paper, we present results from a comprehensive examination of the planimetric
accuracy of various Ikonos 1-m panchromatic image datasets. This analysis includes both the
Geo and Precision imagery from multiple test sites. The impact of DEM resolution and error on
the planimetric accuracy of the resulting orthoimages is also evaluated. In addition, the
planimetric accuracy of USGS DOQQs and other custom NAPP-based orthoimages are
evaluated for comparative purposes. The results are used to assess the cost/benefit of various
digital basemap sources for adoption and use in local government GIS applications.
3. Methods and data sets
3.1 Study areas and image data
Table 1 provides a summary of the study area and Ikonos image characteristics for the three
test sites used in this study. The first study area was in Southern Boone County (SBC), Missouri
4
just south of the City of Columbia (population 70,000). The study area is about 220 km2 (14 x
16 km) and is part of the US 63 highway corridor connecting Columbia, MO and the state capital
of Jefferson City, MO. The study area contains the small town of Ashland (population 1,500).
This is an environmentally sensitive area containing several state parks. The area is under
pressure from urban expansion (primarily single family housing developments) pushing south
out of Columbia and north out of Ashland. The second study area is located in a rapidly growing
suburban area in St. Charles County (SCC), Missouri. SCC is a bedroom community near St.
Louis, MO and is one of the fastest growing counties in the Midwest. The final study area is
located in an urban area of Springfield, MO. This area has been designated as an Urban
Validation Site (UVS) for testing and evaluation of a large variety of remote sensing datasets for
urban applications (Hipple and Daugherty 2000). The SCC and UVS study locations cover areas
of 49 km2 (7 x 7 km) and 21 km2 (3 x 7 km), respectively.
A Carterra Geo Ikonos panchromatic image was acquired for the SBC study area on
4/30/2000 with a nominal viewing angle of 37° (elevation angle measured with respect to
satellite nadir). The Carterra Geo products are not corrected for terrain distortions and
consequently these have the poorest planar accuracy of all products available from Space
Imaging (SI). The worst-case planar accuracy of the Carterra Geo product is 50 m CE90 over
non-mountainous terrain (Space Imaging 2000). For both the SCC and UVS study areas,
Precision Plus Ikonos orthoimages were obtained under the NASA Phase II Scientific Databuy
agreement (NASA 2001) with Space Imaging. Under this agreement, the planar accuracy of
these products is 2 m CE90 compared to the 4 m CE90 accuracy of the Carterra Precision
product that is commercially available from Space Imaging. In addition, Carterra Geo Ikonos
images were also obtained for each test site so that these could be independently processed to
produce orthoimage basemaps. For the SCC study area, the georeferenced Ikonos image was
acquired on 2/29/2000 at a viewing angle of 18°. This is the same source image that what used
by SI to produce the Precision Plus orthoimage under the NASA Phase II Databuy agreement.
For the UVS study area, the SI product was acquired on 3/28/2000 at a viewing angle of 12°,
whereas the georeferenced Ikonos image was acquired separately on 9/17/2000 at a viewing
angle of 20°.
3.2 Digital Elevation Models
A Digital Elevation Model (DEM) of some form is needed to process georeferenced image
data (satellite or airborne) to remove planar distortions in the image caused by terrain variations.
This process is called orthorectification and is required to produce image basemaps with a high
degree of positional accuracy. We created custom DEMs with various horizontal resolutions and
vertical accuracies by stereo-processing aerial photos for all three study areas. For each study
area, B/W film positives were obtained from the NAPP archive at a cost of only $10/photo. The
photos were acquired by the NAPP program in quasi leaf-off conditions in the spring of 1996 for
all three locations. The overlap in the photos is approximately 60% N/S and 30% E/W (Light
1993). Stereo-coverage from 6-10 photos, depending on site location, was needed to produce the
DEMs. The B/W film positives were precision scanned at 1200 dpi (0.85 m pixel size) by a third
party vendor at a total cost of $200 ($20/photo). Precision scanning is required to preserve the
geometric integrity of the B/W photos.
5
A rapid-static differential GPS survey was conducted in each study area to obtain anywhere
from 35-45 ground control points (GCPs) for DEM registration/generation. In addition,
kinematic GPS data were collected between GCPs to generate independent check points (ICPs)
for DEM validation. The accuracy of the rapid-static GCPs was estimated by comparing GPS
coordinate solutions with known positions at multiple monument sites. The accuracy of the
GCPs was found to be 3-5 cm RMS for all three coordinates (x, y, and z). The vertical accuracy
of the kinematic GPS points, used as ICPs for DEM validation, was found to be 10 cm RMS
from crossing-point comparisons following procedures described in Davis and Wang (2001).
The NAPP photos were processed using commercially available software for DEM
extraction (Apex v7.0 - PCI Geomatics 2001). Thus, the results presented here could be easily
reproduced by other parties (e.g. third-party consulting companies, GIS specialists in city/count
governments, etc.). The GCPs are used to generate a highly accurate coordinate reference frame
(CRF) for triangulation and registration of the NAPP photos. DEMs with 3, 10, and 30 m
horizontal resolutions were extracted via automated stereo-correlation processing for the SBC
study area. DEMs with 10 m and 30 m resolution were extracted for the SCC and UVS study
areas. A more complete description of the procedures used here for DEM generation from
stereo-NAPP photographs can be found in Davis and Wang (2001). DEMs with different
horizontal resolutions were created to evaluate the impact of DEM resolution, and the
corresponding vertical accuracy, on the planimetric accuracy of the orthorectified Ikonos images.
A 3-m resolution DEM was not created for the SCC and UVS study areas because of the poor
planimetric accuracy results obtained for the Ikonos images in the SBC study area (Section 4).
In addition to the custom NAPP-based DEMs, 30-m resolution Level 1 DEMs were obtained
from the USGS for each of the study areas. This was done to demonstrate the utility of the
USGS DEMs for generation of orthoimage basemaps. Since the USGS DEMs are publicly
available, these represent an important no-cost elevation data source with easy access for use
when no other elevation data are readily available.
ICP datasets were derived from the kinematic GPS data to assess the vertical accuracies of
the various DEMs in each study area. For each study area, the ICP dataset was a small subset of
the kinematic GPS data. The kinematic GPS data were processed first by eliminating lowaccuracy kinematic positions (due to obstructions caused by trees and buildings during the
survey). In addition, the ICP data were selected: a) for uniform distribution throughout the study
area, and b) for a minimum separation between ICPs of 10 m. The ICP datasets contained
anywhere from 3000-8000 check points depending on the study area.
The RMS vertical accuracy estimated using the ICP datasets for the various DEMs are shown
in table 2. The results show that the RMS vertical accuracies of the custom NAPP DEMs varied
between 1.2-3.0 m. The RMS vertical accuracy of the 3 m NAPP DEM in the SBC study area is
significantly larger than the 10-m and 30-m resolution NAPP DEMs because it becomes more
difficult for the automated stereo-correlation software to precisely determine the location of the
matching pixel pairs in the stereo images as the DEM resolution increases (Davis and Wang
2001). The RMS vertical accuracies of the USGS DEMs were 1.6 m and 1.7 m for the SBC and
UVS study areas, respectively. This is significantly better than the nominal (i.e. target) 3-5 m
vertical accuracy for the USGS DEMs. However, the 16.8 m RMS vertical accuracy of the
6
USGS DEM in the SCC study area is greater than the worst-case specification of 15 m (USGS
1997). Nevertheless, the majority of the DEMs used in this study have RMS vertical accuracies
between 1.2-2.3 m. These accuracies have been determined using thousands of highly accurate
ICPs derived from kinematic GPS surveys. These RMS vertical accuracies are very good and
the corresponding DEMs are appropriate for orthorectification of the georeferenced Ikonos
panchromatic imagery.
3.3 Orthorectification
Digital orthorectified image basemaps were created and/or evaluated using a variety of image
data sources and DEMs. As discussed in Section 3.1, Precision Plus orthorectified Ikonos
images with a specified 2-m CE90 planimetric accuracy were obtained from Space Imaging, via
the NASA Phase II Databuy, for the SCC and UVS study areas. In addition, the USGS DOQQs
were also obtained for all three study areas with a worst-case accuracy of 10 m CE90 (USGS
1996). Orthoimage basemaps were produced from the NAPP and Ikonos Carterra Geo images
using commercially available software (OrthoEngine v7.0 – PCI Geomatics 2001). Thus, these
results could easily reproduced by third parties as needed. This is an important issue for
widespread adoption by the private sector and/or local government entities. Proprietary
algorithms and non-commercial software are in impediment for widespread development and
application of remote-sensing information products.
First, the custom NAPP DEMs with various horizontal resolutions (table 2) were used to
orthorectify the digitally-scanned NAPP aerial photos to produce NAPP digital basemaps for all
three study areas. In this way the effect of DEM resolution and accuracy could be evaluated in
terms of the impact on the visual quality and planar accuracy of the resulting NAPP orthoimage
basemaps. Note that these digital basemaps are produced from the same NAPP data used by the
USGS to produce the DOQQ products.
Next, the low-precision Ikonos Carterra Geo images were orthorectified using a variety of
techniques and DEMs for each of the three study areas. In addition to the DEM, the
orthorectification process requires GCP input data to develop corrections for planar distortions in
the imagery resulting from the Ikonos sensor viewing geometry. There are several methods
available for developing these corrections. These include the simple polynomial (SP), the
rational function polynomial (RFP), and the more rigorous sensor model (SM). The SP method
was not used as this is known to produce very poor results compared to the other two methods
(Toutin and Cheng 2000).
The RFP method uses a ratio of polynomial transformations that take into account ground
elevation in addition to horizontal location to correct for planar distortions in the imagery. The
RFP method does not attempt to model and/or estimate the satellite sensor viewing geometry.
Instead the RFP method uses the GCP data to develop local estimates for the planar distortion
caused by the viewing geometry. Since the RFP method relies exclusively upon the polynomial
transformations derived from GCP data, the RFP corrections are only valid for areas in the
vicinity of the GCP. As a result, planimetric distortions are not entirely eliminated between GCP
locations. Thus, the RFP method is best suited for small areas with gentle terrain and a large
number of GCPs.
7
The SM method uses the GCP data to develop a rigorous model of the sensor viewing
geometry (Toutin 1995) that is valid for the entire image scene. Even though the detailed sensor
information for Ikonos has not been released by Space Imaging, a valid SM solution can be
obtained for Ikonos using GCP input and basic information from the image metadata files
(Toutin and Cheng 2000). A distinct advantage of the SM solution is that only a limited number
of ground control points (4-6) are required to develop a model that is valid for the entire image
scene.
Both the RFP and SM methods were used to generate different Coordinate Reference Frame
(CRF) solutions for each Ikonos Carterra Geo image in a given study area. A total of 10 GCPs
obtained from the rapid-static GPS survey (section 3.2), uniformly distributed over the study
area, were used to develop the RFP and SM solutions. The remaining rapid-static GPS points in
each study area where then set aside to be used as Independent Check Points (ICPs) for
validation of the planimetric accuracy of the resulting orthoimage basemaps. Figure 2 shows the
spatial distribution of the GCPs and ICPs in the Ikonos basemap that was produced for the SBC
study area using the 10-m NAPP DEM and the SM coordinate solution. Table 3 summarizes the
RMS residual errors for the RFP and SM solutions obtained from least-squares bundle
adjustment of the GCP input data. The same 10 GCP points were used in both the RFP and SM
solutions for each study area. For all three study areas the CRF RMS errors are excellent and are
less than one-half pixel (< 0.5 m). This is due to: 1) the high horizontal and vertical positional
accuracies of the GCP input data, 2) the a priori selection of the GCP locations at point targets in
the imagery with high contrast and visual clarity, and 3) the ability of the software operator to
select the GCP locations in the imagery with sub-pixel accuracy.
After the two CRF solutions were obtained for each Ikonos Carterra Geo image, the images
were then orthorectified using various DEMs summarize in table 2 for each study area. The
planimetric quality and accuracy of the DOQQ, NAPP, and all the various Ikonos orthoimage
basemaps were then evaluated. The planimetric quality was subjectively measured by examining
the linearity of known linear features (primarily roads and buildings) in the imagery. The
planimetric accuracy was assessed using ICP datasets that contained 20, 24, and 35 points
uniformly distributed (e.g. figure 2) throughout the SBC, SCC, and UVS study areas,
respectively. As noted previously, the ICP coordinates were obtained from rapid-static GPS
survey data and had RMS x and y accuracies between 3-5 cm (section 3.2). The ICPs were
selected a priori at locations in the imagery that were sharp and distinct point features. This
facilitated the identification of these points in the orthoimage basemaps and subsequent
comparison with the known GPS-based ICP locations. The ICPs are distinct from the GCPs used
for the RFP and SM solutions for the Ikonos Carterra Geo images (e.g. figure 2).
The RMS Radial Error (RMSRE) and the Circular Error @ 90% probability (CE90) were
calculated for each orthoimage basemap using the ICP datasets. The RMSRE is given by
RMSRE = RMS x2 + RMS y2
(1)
where RMSx and RMSy are the RMS errors in the x and y directions, respectively. The RMS
error in each coordinate is computed using
8
N
RMS =
∑ (P
i =1
ICPi
− PIMAGEi ) 2
N −1
(2)
where PICP and PIMAGE are the x or y positions obtained from the ICPs and the image locations,
respectively. The CE90 is estimated by rank ordering the radial errors from smallest to largest
and selecting the radial error corresponding to the 90th percentile, i.e. the abscissa value of the
90% point of the Cumulative Distribution Function (CDF) of the radial (circular) error is used.
For example, in an ICP dataset of N = 20 points, the 18th (0.9 × 20) radial error value in the rank
ordered dataset is used to estimate CE90. For a purely random error distribution in both x and y
coordinates, the relationship between the circular error @ 90% and the RMSRE is CE90 = 1.54 ×
RMSRE. The actual relationship will vary slightly from this due to small bias errors in the x and
y directions.
4. Results and discussion
4.1 DOQQ and NAPP basemaps
The planimetric quality and accuracy of the DOQQ and custom NAPP digital basemaps are
summarized in tables 4-6. The USGS DOQQ basemaps had CE90 values between 2.7 m and 3.4
m. Thus, the DOQQ basemaps in this study had planar accuracies that are suitable for most local
government applications where a nominal accuracy of 3-4 m CE90 is required. Note that the
specified worst-case planar accuracy of the DOQQs is 10 m CE90 (USGS 1996), so the results
presented here certainly may not be valid for all DOQQs. However, we believe that the 10-m
CE90 specification is indeed a worst-case value and that many DOQQs will have planar
accuracies substantially better than this, just as the case for the three test sites in this study.
The custom NAPP basemaps, which are derived from the same aerial photographs as the
USGS DOQQs, had CE90 values that ranged between 0.9 m and 3.1 m depending on study area
and the DEM source used for orthorectification. For the SBC study area, the lowest and highest
accuracy custom NAPP basemaps were orthorectified using the 3-m and 30-m NAPP DEMs
respectively. We attribute that higher planar accuracy of the NAPP basemap orthorectified using
the lower resolution 30-m DEM to the better vertical accuracy and coarser spatial resolution. As
noted previously, the 3-m NAPP DEM had an RMS vertical accuracy of 3.0 m while the 30-m
NAPP DEM had an RMS vertical accuracy of 2.2 m (table 2). The coarser resolution of the 30m DEM in effect represents a spatial average of the elevation data relative to the 3-m NAPP
DEM. Abrupt (erroneous) elevation discontinuities present in the 3-m NAPP DEM are
smoothed out in the coarser resolution 30-m NAPP DEM.
Figure 3 shows a comparison of how well the DOQQ and custom NAPP basemaps preserved
the linearity of roads in the SBC study area. Subjective rankings of poor, fair, good, and
excellent are provided in tables 4-6 to describe the planimetric quality of the various image
basemaps based upon the preservation, or lack thereof, of known linear features. Strong
distortions are present in the NAPP basemap derived from the 3-m NAPP DEM. Moderate
distortions are present in the NAPP basemap derived from the 10-m NAPP DEM, whereas no
9
significant distortions occur in the DOQQ and the NAPP basemap derived from the 30-m NAPP
DEM. Note also that the contrast and clarity of the custom NAPP images are slightly better than
the DOQQ.
There is clearly a tradeoff between horizontal resolution and vertical accuracy of a DEM and
its subsequent effect on the planar accuracy and quality of basemaps derived from DEMs via the
orthorectification process. In general, it is desirable to use the highest resolution DEM possible
for orthorectification. This is especially true in urban areas where there are substantial
topographic variations over short spatial scales (e.g. buildings). In these areas, high resolution
DEMs are needed to preserve the planar accuracy of orthorectified image features that are often
adjacent to each other and have substantially different elevations (e.g. roads and buildings).
However, the vertical accuracy of DEMs derived from stereo-correlation procedures usually
degrades as the resolution increases (assuming the source image resolution is the same). Thus,
the planimetric accuracy and visual quality tends to degrade as the DEM resolution increases
because the vertical accuracy worsens. In more rural areas with less elevation variability over
short spatial scales, it is preferable to utilize moderate resolution DEMs that maintain an
acceptable vertical accuracy. The results in table 4-6 show, in general, that the custom NAPP
basemap derived from the 30-m NAPP DEM provides the best planar accuracies and quality
amongst the NAPP/DOQQ basemaps. In a more urban setting, a 30-m DEM may not be the
best choice for orthorectification. Obviously we desire the highest resolution and highest
accuracy DEM for orthorectification. However, in most situations one must critically evaluate
both the resolution and vertical accuracy of the DEM to determine the proper choice for digital
basemap generation via orthorectification.
The planimetric accuracies for the custom NAPP basemaps derived using the 30-m NAPP
DEM were around 1.5 m CE90 for all three study areas (tables 4-6). This is significantly better
the DOQQ planar accuracies for any of the areas studied here. Thus, the raw aerial photographs
in the NAPP archives can be exploited to produce orthoimage basemaps with excellent planar
accuracies when the planar accuracies of the DOQQs are found to be insufficient or when
specific applications require planar accuracies on the order of 1-2 m CE90. Note that DEMs
with suitable horizontal resolutions (10-30 m) and RMS vertical accuracies (~ 2 m) are required
to achieve the 1-2 m CE90 planar accuracy.
4.2 Ikonos Basemaps
The planimetric quality and accuracy of the Ikonos digital basemaps are also summarized in
tables 4-6. The low-precision Ikonos Carterra Geo products had planar accuracies between 9 m
and 24 m CE90. This is significantly better than the 50-m CE90 worst-case specification (Space
Imaging 2000) and is due largely to the small variations in terrain and/or small coverage areas
(table 1). As expected, the largest planar error occurred for the SBC test site which had greater
terrain variations over larger spatial scales than the other two test sites.
The planimetric accuracies of the Ikonos Precision Plus basemaps provided by Space
Imaging were 1.1 m and 1.7 m CE90 for the SCC and UVS study areas, respectively. A 2-m
CE90 planar accuracy was required under the terms of the NASA Phase II Databuy agreement
with Space Imaging. Thus, the Space Imaging products were found to meet the NASA Phase II
10
Databuy specification. This independent assessment therefore confirms that Space Imaging
precision orthoimages can produce planar accuracies on the order of 2-m CE90 under suitable
conditions (e.g. viewing angle, GCP input, high-resolution DEM, etc.). The planar accuracy of
these products approaches the 1-m pixel size of the panchromatic Ikonos imagery and is likely
the best that can be achieved for these data. The 1-2 m CE90 planar accuracy of these
orthoimages combined with the overall clarity and contrast of the Ikonos panchromatic imagery
represents an extremely high quality product for digital basemap applications. The only factor
limiting the widespread adoption of the Space Imaging precision orthoimages by local
governments is the higher cost of the Carterra Precision products ($63/km2) relative to the
Carterra Geo products ($21/km2).
The custom Ikonos orthoimages generated by orthorectifying the Ikonos Carterra Geo source
images using the 10-m and 30-m NAPP DEMs and the RFP and SM coordinate solutions yielded
planimetric accuracies that ranged from 2.7-4.6 m CE90, 1.8-4.4 m CE90, and 1.1-2.0 m CE90
for the SBC, SCC, and UVS study areas respectively (tables 4-6). In all study areas, the RFP
solutions appear to produce slightly better planar accuracies relative to the corresponding SM
solutions for the same source DEM. For example, the RFP basemaps for the SBC study area had
CE90 values ranging from 2.7-3.1 m CE90, while the SM basemaps had CE90 values ranging
from 3.2-4.6 m.
This differs substantially from the results obtained by Toutin and Cheng (2000) who found
that the SM solution was far better than the RFP solution (RMSRE ~ 1.8 m for SM vs. 5.6 m for
RFP) for one Ikonos test image of Richmond Hill, Ontario. Just as was the case in the Toutin
and Cheng (2000) study, only ICPs were used (not GCPs) for assessing the planar accuracy, and
these were uniformly distributed throughout the study area (e.g. figure 2). These issues are
important because the RFP method corrects locally at the GCP. The accuracy assessment could
be significantly biased if only GCPs are used in computing positional errors or if the ICPs are
located in the vicinity of the GCPs. Since this is not the case in our study, this cannot
satisfactorily explain the difference between the RFP and SM results obtained here and those
from the Toutin and Cheng (2000) study. One possible explanation for this may be due to the
number of GCPs used in the RFP coordinate solutions. We used 10 GCPs for all RFP solutions
compared to 7 GCPs for the Toutin and Cheng (2000) study. As a result, a higher order RFP
solution was obtained for our study and more local areas are adjusted as a result. This could
account for the observed differences between the RFP and SM planar accuracies between the two
studies.
However, even though all the ICP datasets indicate that the Ikonos RFP basemaps have better
planar accuracies than the Ikonos SM basemaps, visual examination shows that substantial linear
distortions (i.e. warping) may occur for the RFP basemaps. Figure 4 presents a comparison of
the planimetric quality of the RFP and SM basemaps for the SBC study area for various DEM
data sources. This is the same location as that portrayed in figure 3. The comparison in figure 4
shows that the RFP method can produce substantial linear distortions that render these basemaps
less desirable relative to the corresponding SM basemaps. This difference is especially
noticeable for the basemaps derived from the higher resolution DEM (10-m NAPP). This occurs
even though the ICP dataset indicates the planar accuracies are slightly worse for the SM
solution (table 4).
11
It is very likely that distortions in the RFP basemaps, caused by the local nature of the RFP
solution, are not present in the vicinity of the ICPs, and this therefore accounts for the apparently
higher planar accuracy relative to the SM basemaps. In reality, the Ikonos SM basemaps have
better planar accuracies relative to their RFP counterparts for the SBC study area when the same
source DEM is used. The significant differences in the preservation of the linear features
resulted in higher planimetric quality rankings for the SM solutions compared to the RFP
solutions for the SBC study area. For the SCC study area, the planimetric quality differences
between the SM and RFP solutions were still noticeable but were not as great as the differences
for the SBC study area. For the UVS study area, there were no noticeable differences in the
planimetric quality between the SM and RFP solutions when the same source DEM is used. The
planimetric quality rankings for the SCC and UVS study areas (tables 5 and 6) reflect these
observations.
The differences in the planimetric quality between the SM and RFP solutions are significant,
noticeable, and non-existent for the SBC, SCC, and UVS study areas, respectively. This
variation is directly related to the size differences of the study areas (table 1). Recall that 10
GCPs were used for the SM and RFP solutions for each study area. Since the RFP method only
corrects for planimetric distortions locally at each GCP, the GCP density per unit area is an
important characteristic affecting the quality of the RFP basemaps. The GCP density per km2
was 0.48, 0.20, and 0.04 for the UVS, SCC, and SBC study areas respectively. The factor of 10
difference in the GCP density between the UVS and SBC study is responsible for the substantial
differences in the planimetric quality between the two study areas. Since the RFP method only
corrects locally at each GCP for planimetric distortions, it is only suitable for small study areas
or areas were a large number of highly accurate GCPs are available. The results here suggest
that a GCP density of roughly 0.4/km2 is required for the RFP method to produce no noticeable
linear distortions in the orthorectified image basemap. Since a typical Ikonos image scene is 11
x 11 km (121 km2), this would require 48 GCPs to produce an RFP solution without noticeable
linear distortions in the image area. This is clearly impractical for most applications, and thus
the RFP method should only be utilized for small study areas.
A distinct advantage of the SM solution is that only a limited number of ground control
points are required to develop a model that is valid for the entire image scene. For the SM
solutions, the custom Ikonos orthoimages generated by orthorectifying the Ikonos Carterra Geo
source images using the 10-m and 30-m NAPP DEMs produced planimetric accuracies that
ranged between 1.9 m and 4.6 m CE90 for the three study areas. The best planar accuracies for
the SM solutions using the NAPP DEMs were 2.3 m and 1.9 m CE90 for the SCC and UVS
study areas, respectively. These are significantly better than the 3.2 m CE90 planar accuracy
obtained in the SBC study area from the SM solution and the 10-m NAPP DEM. We attribute
this directly to the smaller viewing angles (18° and 20°) of the Ikonos Geo images for the SCC
and UVS study areas relative to the 37° viewing angle for the SBC study area (see table 1).
Small viewing angles are better suited for accurate orthoimage generation because horizontal
pixel displacements caused by topographic variations and the corresponding susceptence to
DEM error will be minimized. For the commercial SI Ikonos orthoimage products, the SI
technical specifications require a nominal viewing angle <15° for production of precision
orthoimages with 4 m CE90 (Space Imaging 2000). Thus, generation of the highest precision
12
orthoimage basemaps from georeferenced Ikonos imagery require acquisition of the image from
a small viewing angle. However, we note that a 3.2 m CE90 accuracy was obtained for the SBC
study area from the georeferenced Ikonos source image with a 37° viewing angle. Thus, it is
possible to achieve planar accuracies comparable to the commercial SI Carterra Precision
product with 4 m CE90 even when the source image has a larger than nominal viewing angle.
The results presented here clearly demonstrate that Ikonos orthoimage basemaps with planar
accuracies of ~2.0 m CE90 can be obtained using the low cost, low precision Ikonos
georeferenced imagery with viewing angles ≤ 20° (e.g. SCC and UVS test sites). This accuracy
level was achieved using: 1) commercially available orthorectification software, 2) a small
number of GCPs (e.g. 10) to produce a valid sensor model solution, and 3) NAPP DEMs with
horizontal resolutions of 10-30 m and RMS vertical accuracies on the order of 2 m. This
effectively demonstrates that low cost, low precision georeferenced Ikonos imagery can be used
by independent parties (e.g. no affiliation with SI) to produce orthorectified digital basemaps
with planar accuracies on the order of 2 m CE90. This is comparable to the SI/NASA Databuy
products evaluated here and better than the highest precision orthoimage products commercially
available from SI (Carterra Precision, 4 m CE90). From the standpoint of local governments, the
most important benefit of this approach is the cost savings, as the georeferenced Ikonos image
products are available for a fraction of the cost of the precision orthoimage products ($21/km2 vs.
$63/km2).
The results presented in the preceding paragraphs were generated using custom NAPP DEMs
for performing the Ikonos image orthorectification. This can be an impediment for local
governments who do not already possess sufficiently accurate digital elevation data and who do
not have the technical expertise to generate custom NAPP DEMs. Since the USGS 30-m DEMs
are readily available throughout most of the U.S., these existing DEMs are a potential solution to
this problem. The planar accuracy of the Ikonos SM basemap generated using the USGS DEM
was 2.0 m CE90 (table 6) for the UVS study area. For the SBC study area, the planar accuracy
of the Ikonos SM basemap generated using the USGS DEM was 4.6 m CE90 (table 4). Thus, it
is clear that the USGS DEMs can be used to orthorectify georeferenced Ikonos data to yield
imagery with planar accuracies that are sufficient for local government basemap applications.
However, we note that the resulting planar accuracy will of course depend on the vertical
accuracy of the USGS DEM. Planar accuracies of 2-4 m CE90 will not automatically be
obtained using USGS DEMs. For the SBC and UVS study areas, the RMS vertical accuracies of
the 30-m USGS DEMs were found to be 1.6 m and 1.7 m, respectively, and these produced
Ikonos SM basemaps with 4.6 m and 2.0 m CE90 planar accuracies. However, for the SCC
study area, the Ikonos SM basemap derived from the USGS DEM yielded a planar accuracy of
only 7.0 m CE90. This comparatively poor result is due to the poor RMS vertical accuracy (16.6
m) of the particular USGS DEM available for the SCC study area. Thus, it is essential that some
validation of the vertical accuracy of USGS DEMs be done prior to using them for production of
Ikonos orthoimage basemaps.
Finally, the planimetric accuracy results presented in tables 4-6 were estimated using ICP
datasets, derived from rapid-static GPS survey, that contained anywhere between 20-35 points.
Even though the ICPs were uniformly distributed throughout each study area, this limited
13
amount of checkpoints will imperfectly estimate the RMS and CE90 accuracies. Due to
sampling variability alone, the RMS and CE90 values reported in tables 4-6 are accurate to about
20-30%. To further validate the utility of the SM solutions for orthorectification of Ikonos
georeferenced imagery, we collected 230 rapid-static GPS positions in a 630-km2 study area (19
x 33 km) located in Boone County, Missouri. Two georeferenced Ikonos images were acquired
in April 2000 covering the complete study area. The viewing angles of the two images were 27°
and 37° respectively. A set of 10 uniformly distributed GCPs was used to generate the SM
solutions for each image. USGS DEMs with 30-m horizontal resolution were used to
orthorectify the Ikonos images, and the two orthorectified images were combined into a single
mosaic. An ICP datasets comprised of 210 rapid-static GPS positions was used to validate the
planimetric accuracy of the orthomosaic. Figure 5 shows a scatter plot of planimetric errors in
Ikonos orthomosaic. The distribution of the planimetric errors is clearly random and the
corresponding planimetric accuracies were 3.1 m RMSRE and 4.5 m CE90. These are consistent
with the SBC results in table 4 which utilized an Ikonos image with a large viewing angle as well
(27°).
This further demonstrates that the low cost, low accuracy georeferenced Ikonos image
products can be orthorectified by third parties using existing DEM data sources and limited
amounts of ground control to produce digital image basemaps with planimetric accuracies
suitable for local government applications. Planimetric accuracies on the order of 2.0 m CE90
can be obtained when the acquisition viewing angle is ≤ 20°. The comparatively lower cost of
the georeferenced Ikonos imagery makes these very attractive for use by local governments for
developing up-to-date digital image basemaps for incorporation into GIS systems.
5. Summary and conclusions
In this study we presented a detailed assessment of the planimetric accuracy of highresolution (1 m) panchromatic Ikonos orthoimage products for three different test sites located in
the State of Missouri. The main objective of this study was to evaluate the potential of Ikonos
orthoimage products for use as digital image basemaps for use in local government GIS systems.
For maximum utility and adoption in a wide variety of local government planning and
management applications, a planimetric accuracy of 3-4 m CE90 is considered nominal. The
planimetric accuracy of USGS DOQQs and custom orthoimages derived from NAPP aerial
photography were also evaluated for comparative purposes. A variety of DEM data sources with
different horizontal resolutions and vertical accuracies were used to determine their affect on the
planimetric accuracy of orthoimage products derived from the DEMs. Key findings from this
study are:
(1) The USGS DOQQs had planimetric accuracies on the order of 3 m CE90 for the test sites
in this study. This is substantially better than the 10 m CE90 specified by the USGS. As a
result, we believe that many DOQQ products are likely to meet the desired accuracy of 3-4 m
CE90 for local government applications.
(2) Custom orthoimages derived from NAPP aerial photographs were produced and were
found to typically have planimetric accuracies on the order of 1-2 m CE90. These accuracies
14
are significantly better than the DOQQs produced by the USGS from the same NAPP
imagery. The raw aerial photographs in the NAPP archives can be used to produce basemaps
with excellent planimetric accuracies when the accuracy of the DOQQs is found to be
insufficient.
(3) DEMs with horizontal resolutions of 10-30 m and RMS vertical accuracies of 2 m are
required to achieve planimetric accuracies on the order of 2-4 m CE90 in orthoimage
products derived from the DEMs.
(4) The planimetric accuracies of Ikonos orthoimages acquired under a Space
Imaging/NASA Databuy agreement were found to be 1.1 m and 1.7 m CE90 at two different
test sites. The terms of the NASA Databuy agreement required a 2 m CE90 planimetric
accuracy. Our independent assessment confirms that the Space Imaging products met this
NASA Databuy specification.
(5) Low precision georeferenced Ikonos image products (Carterra Geo, 50 m CE90) were
orthorectified using third party commercially available software and various custom NAPP
DEMs. The planimetric accuracies of the resulting Ikonos orthoimages were found to vary
between 2-4 m CE90.
(6) Existing USGS DEMs were used to orthorectify georeferenced Ikonos image products.
The resulting Ikonos orthoimages were found to have planimetric accuracies from 2-7 m
CE90. Planimetric accuracies of 2-4 m CE90 can be obtained from georeferenced Ikonos
using USGS DEMs with RMS vertical accuracies on the order of 2 m. An image acquisition
viewing angle for the Ikonos image would likely need to be ≤ 20° to obtain the highest
planimetric accuracies (e.g. 2 m CE90).
The most important results from this study are with respect to cost effective acquisition of
orthoimagery suitable for use as digital image basemaps in local government GIS applications.
The planimetric accuracies of the USGS DOQQs and custom NAPP orthoimages were found to
meet or exceed the desired basemap accuracy of 3-4 m CE90. These orthoimage products
represent very low cost GIS basemap data sources. The main disadvantage of these products is
that they are based on NAPP aerial photographs that are acquired only once every 5-6 years.
Thus, these images are often outdated in urban areas that experience even moderate growth.
This therefore limits the use of these orthoimages in many local government planning and
management applications (tax assessment, E-911, etc.).
The recent launch of the Ikonos satellite has opened up a new area for acquiring up-to-date
high-resolution panchromatic digital imagery for use as digital orthoimage basemaps. The
Ikonos panchromatic imagery has a nominal pixel size of 1 m and an 11-bit information content.
The 11-bit imagery provides image contrast and clarity that is far superior to the DOQQs. The
Ikonos panchromatic image product with the lowest horizontal precision (Carterra Geo) costs
$21/km2 and has a planar accuracy of only 50 m (CE90). While this product is affordable for
most local government entities, the planar accuracy of 50 m renders this product unusable for
basemap applications. The results presented here independently demonstrate that the Carterra
Geo Ikonos products can be orthorectified using the readily available USGS 30 m DEMs to
15
produce digital basemaps with planimetric accuracies on the order of 2-4 m CE90. This level of
accuracy is comparable to the highest accuracy Ikonos image products (Carterra Precision, 4 m
CE90) commercially available from Space Imaging at a cost of $63/km2 (3x more expensive than
Carterra Geo). A cost for complete coverage of a typical county (2400 km2) within the State of
Missouri would be $50,400 and $151,200 for the Geo and Precision Space Imaging products,
respectively. The latter figure is well beyond the budgetary capability of most local
governments. The former figure is cost-competitive with traditional aerial photography surveys
for the same resolution and coverage. Thus, the approach demonstrated here can be used to
develop up-to-date, cost effective orthoimages from Ikonos Carterra Geo products that yield
planimetric accuracies suitable for use as digital image basemaps by local governments.
Acknowledgements
The SBC work was supported by the Raytheon Synergy program under subcontract
#012100MJ-3 from NASA. The SCC and UVS work was supported by NASA Stennis Space
Center under contract # NAG13-99014 and by the Institute for the development of Commercial
REmote Sensing Technologies (ICREST) at the University of Missouri-Columbia.
16
References
Davis, C.H. and X. Wang, 2001, “High resolution DEMs for urban applications from NAPP
photography,” Photogrammetric Engineering & Remote Sensing, Vol. 67, No. 5, pp. 585-592.
Hipple, J. and D. Daugherty, 2000, “Urban validation site for testing impervious surface models
derived from remotely sensed imagery,” Proceedings of International Geoscience and Remote
Sensing Symposium, Vol. 7, pp. 2885-2889, Honolulu, Hawaii, 24 - 28 July, 2000.
Light, D.L, 1993, “The National Aerial Photography Program as a Geographic Information
System Resource,” Photogrammetric Engineering & Remote Sensing, Vol. 59, No. 1, pp. 61-65.
PCI Geomatics, 2001, OrthoEngine V7.0, URL: http://www.orthoengine.com, PCI Geomatics,
Ontario, Canada.
Space Imaging, 2000, “Space Imaging Catalog of Products and Services,” Volume 1 Supplement
(55p.), Thornton, CO.
United States Geological Survey, 1996, “Standards for Digital Orthophotos, Part 1: General (9
p.); Part 2: Specifications (37 p.),” Department of the Interior, Washington, DC.
United States Geological Survey, 1997, “Standards for Digital Elevation Models, Part 1: General
(17 p.); Part 2: Specifications (70 p.); Part 3: Quality Control (10 p.),” Department of the
Interior, Washington, DC.
NASA, 2001, Scientific Databuy – Phase II, http://www.crsp.ssc.nasa.gov/scripts/databuy,
NASA Stennis Space Center, Stennis, MS.
Toutin, T. and P. Cheng, 2000, “Demystification of IKONOS,” Earth Observation Magazine,
Vol. 9, No.7.
17
List of Tables
Table 1. Study area and Ikonos image characteristics.
Test
Site
Area
Type
Test
Area
(km)
Terrain
Variation
(m)
Product Type
Acquisition
Viewing
Angle
SBC
Rural
14 x 16
160-290
Geo1
4/30/2000
37°
SCC
Suburban
7x7
140-200
Geo
Prec Plus2
2/29/2000
2/29/2000
18°
18°
UVS
Urban
3x7
330-430
Geo
Prec Plus
9/17/2000
3/28/2000
20°
12°
1
2
Ikonos Panchromatic Images
Geo = Carterra Geo product from Space Imaging with 50 m CE90
Prec Plus = Precision Plus product from Space Imaging with 2 m CE90 for NASA Databuy contract
Table 2. DEM horizontal resolutions and vertical accuracies for the three study areas.
Test Site
SBC
SCC
UVS
DEM
NAPP
NAPP
NAPP
USGS
NAPP
NAPP
USGS
NAPP
NAPP
USGS
XY
RMS Z
Resolution Accuracy
(m)
(m)
3
3.0
10
2.3
30
2.2
30
1.6
10
1.2
30
1.3
30
16.8
10
2.1
30
2.3
30
1.7
Table 3. RMS residual errors from the Rational Function Polynomial (RFP)
and Sensor Model (SM) solutions for Ikonos Geo imagery.
RFP Solution
SM Solution
Test Site RMS X RMS Y RMSRE RMS X RMS Y RMSRE
(m)
(m)
(m)
(m)
(m)
(m)
SBC
0.26
0.12
0.29
0.33
0.33
0.47
SCC
0.26
0.19
0.32
0.25
0.23
0.34
UVS
0.23
0.22
0.32
0.27
0.17
0.32
18
Table 4. Planimetric accuracy of orthoimage basemaps for Southern Boone County (SBC) study area.
Image
Data
Source
DOQQ
NAPP
NAPP
NAPP
Ikonos – Geo
Ikonos Ortho – RFP
Ikonos Ortho – SM
Ikonos Ortho – RFP
Ikonos Ortho – SM
Ikonos Ortho – RFP
Ikonos Ortho – SM
DEM
Used
N/A
NAPP 3 m
NAPP 10 m
NAPP 30 m
N/A
NAPP 10 m
NAPP 10 m
NAPP 30 m
NAPP 30 m
USGS 30 m
USGS 30 m
RMS
Radial
Error (m)
2.0
1.9
1.8
1.0
19.5
1.8
2.2
1.6
2.5
2.1
2.9
Circular
Error @ 90%
(m)
3.4
3.0
3.1
1.5
23.7
2.9
3.2
3.1
4.6
2.7
4.6
Planimetric
Quality
Good
Poor
Fair
Excellent
Excellent
Poor
Fair
Fair
Good
Good
Excellent
Table 5. Planimetric accuracy of orthoimage basemaps for St. Charles County (SCC) study area.
Image
Data
Source
DOQQ
NAPP
NAPP
Ikonos – Geo
Ikonos – Prec Plus
Ikonos Ortho – RFP
Ikonos Ortho – SM
Ikonos Ortho – RFP
Ikonos Ortho – SM
Ikonos Ortho – RFP
Ikonos Ortho – SM
DEM
Used
N/A
NAPP 10 m
NAPP 30 m
N/A
N/A
NAPP 10 m
NAPP 10 m
NAPP 30 m
NAPP 30 m
USGS 30 m
USGS 30 m
RMS
Radial
Error (m)
1.9
1.3
0.9
5.4
0.9
1.2
1.6
1.7
2.6
5.8
4.5
Circular
Error @ 90%
(m)
2.8
0.9
1.7
9.0
1.1
1.8
2.3
2.9
4.4
8.0
7.0
Planimetric
Quality
Good
Good
Excellent
Excellent
Excellent
Fair
Good
Good
Excellent
Good
Excellent
Table 6. Planimetric accuracy of orthoimage basemaps for Springfield UVS study area.
Image
Data
Source
DOQQ
NAPP
NAPP
Ikonos – Geo
Ikonos – Prec Plus
Ikonos Ortho – RFP
Ikonos Ortho – SM
Ikonos Ortho – RFP
Ikonos Ortho – SM
Ikonos Ortho – RFP
Ikonos Ortho – SM
DEM
Used
N/A
NAPP 10 m
NAPP 30 m
N/A
N/A
NAPP 10 m
NAPP 10 m
NAPP 30 m
NAPP 30 m
USGS 30 m
USGS 30 m
RMS
Radial
Error (m)
1.9
1.2
1.1
11.3
1.1
0.7
1.3
1.1
1.2
0.9
1.1
Circular
Error @ 90%
(m)
2.7
2.0
1.7
13.5
1.7
1.1
1.9
1.7
2.0
1.6
2.0
Planimetric
Quality
Good
Good
Excellent
Excellent
Excellent
Good
Good
Excellent
Excellent
Excellent
Excellent
19
Figures
Figure 1. Vector data layer overlay of parcel boundaries on a sample USGS DOQQ. Note that the parcel
boundaries divide many single-family residential houses due to the poor positional accuracy of the image
basemap.
20
Figure 2. Ikonos 1-m resolution orthoimage for SBC study area produced from low precision
georeferenced Space Imaging product using 10-m NAPP DEM and SM solution. GCPs and
ICPs are shown as red and yellow triangles, respectively, and are uniformly distributed
throughout the study area.
21
(a)
(b)
(c)
(d)
Figure 3. Examples of DOQQ and custom NAPP orthoimage basemaps for SBC study area illustrating
differences in planimetric image quality from different DEM data sources. (a) DOQQ (b) NAPP with 3-m
DEM (c) NAPP with 10-m DEM and (d) NAPP with 30-m DEM. Note the significant linear feature
distortion in the NAPP orthoimages produced from the 3-m and 10-m NAPP DEMs.
22
(a)
(b)
(c)
(d)
(e)
(f)
Figure 4. Examples of Ikonos orthoimage basemaps from SBC study area illustrating differences in
planimetric quality resulting from different DEM data sources and coordinate solution methods. (a) RFP
with 10 m NAPP DEM (b) SM with 10 m NAPP DEM (c) RFP with 30 m NAPP DEM (d) SM with 30 m
NAPP DEM e) RFP with 30 m USGS DEM and f) SM with 30 m USGS DEM.
23
6
CE90 = 4.5 m
4
Y Error (m)
2
0
-2
-4
RMSRE = 3.1 m
-6
-10
-8
-6
-4
-2
0
2
4
6
8
10
X Errror (m)
Figure 5. Scatter plot of planimetric errors in orthorectified Ikonos image produced using 30-m USGS
DEMs and SM coordinate solution. The ICP dataset contains 210 points derived from rapid-static GPS
survey.