International Archives of the Photogrammetry,
QuickBird RMSE X,Y <> view angle
(DTM spacing) dH
$8.84 5 SS x»
e e e e e e e N N N S SO
40 —————— — y 16
3.5 i A 1d =
E 30 ttt 42 8 REFERENCES
> 2.54 r= L4
x 20 | x | S 2 Cheng P., Toutin Th., Zhang Y., Wood M., 2003. QuickBird -
2 10] 3 I | 3 | 1S geometric correction, path and block processing and data
9 ig. / | | E A rit 4 a
os B Hi LI | Í I | I I | a * fusion. EOM , Vol. 12, No 3 , May , 2003 , pp. 24-30
oo MM M RE RR,
Eurimage, 2004, QuickBird products i
QS, 3.35 6, 2,2 S € € $ $5 rte proie homes KE
+ 2S 0A Te http://www.eurimage.com/products/quickbird.html
Ku. lu AA fos "nil A :
q 79a T. oo ^ ^om p accessed 29 March 2004
T uA D Yd, todo p. S, t, ^,
oP its i eM iis , v ; + 3
European Commission, 2003, Guidelines for Quality Checking AB
SITES (model, no. of GCP) ^
of Ortho Imagery.
wm RMSE X C3 RMSE Y -a- view angle http://mars.jrc.it/marspac/CwRS/default.htm [n 1
. ~* . ~ . % ace
Grodecki J., Dial G., 2003. Block adjustment of high resolution d
Figure 5 ickBird RMSE 'ersus image view angle, for ite i i i 7 2 ii
igure + Quic 1d SMS ne y Versus image view angie, TO! satellite images described by rational polynomials. PE&RS,
different sites. The additional attached information concerns: vol.69, No.1, pp. 59-6 Ge
- for I* X axis: applied model (R — RPC, P — parametric), - ; 34 200 à cak
number of GCPs used for the orthorectification, * GCPs Grodecki J, Dial G., 2002. [konos geometric accuracy dim
= validation. IV/ISPRS Commission [/FIEOS 2002
from map, Cont p di
- for 2" X axis: (DTM spacing), dH difference between onierence P racecdimes. The
extreme values of heights (relief) measured on GCPs and Grodecki J.. Dial G., 2001. *IKONOS Geometric Accuracy." er
check points. Proceedings of Joint Workshop of ISPRS Working Groups
1/2, 1/5 and IV/7 on High Resolution Mapping from Space The
6. CONCLUSIONS 2001, University of Hanover, Hanover, Germany, Sept 19- lor
21, 2001 imr
For Ikonos and QuickBird, the results meet the geometric forc
accuracy requirement of 2.5m RMSE;p. There is no observed Jacobsen K., 2002. Geometric aspects of the handling of space
significant difference between the RPC and parametric model images. Pecora 15/ Land Satellite Information IV, [SPRS In t
performance (PCI Geomatica and Socet Set based). Where Commission I, FIEOS Conference Proceedings. sam
ancillary dz y a gooc ity, any increase i or of ; ; , rang
gneillats data was of a good quality, any increase in number of Kay S., Spruyt P., Kyriacos Alexandrou K., 2003. Geometric
GCP's above the recommended level made no difference to The Ji, ^ dif rte ; thes
; , = quality assessment of orthorectified vhr space image data.
final accuracy. Based on the results obtained in the context of PE&RS May dete
this trial, that in respect of using a limited number of GCPs, the I
RPC based approach may be considered as very reasonable and Space Imaging, 2004. Ikonos products.
practical solution applicable for single-image orthorectification http//www.spaceimaging.com/products/ikonos/geo techspec
of VHR satellite images. For the images tested the view angle htm
values up to 20.15? (Ikonos) gave acceptable results. http//www.spaceimaging.com/products/ikonos/geo ortho.h The
Further tests are needed to check to which extent extreme tm (accessed 29 March 2004) the
off-nadir angles can be used for successful orthorectification in io the i ; ; ent
X 2 d d f : "U Aeprion : Toutin Th.. 2003. Geometric Correction of Remotely Sensed fico
this type of applications. Since higher off-nadir angles permit ; a m^ ned ee
d UI, this i aT Lh id Fa Images; Remote Sensing of Forest Environments: Concepts mai
reduced revisit times, this is an important consideration. “ de eo HC ; ; :
Ke S ims is an qp uy . and Case Studies (Chapter 6), edited by: M.A. Wulder, S.E. cha
The study confirmed the importance of the quality of : 3 : bre 7 3.180
; OR 2 N ; 3 7 Franklin , Kluwer Academic Publishers , 2003 , pp. 14 -1 itf
ancillary data. The visible influence of quality of ancillary data
(DEM, GCP’s) on the accuracy caused also the partial masking Toutin Th., R. Chénier, Y. Carbonneau; 2002. 3D models for Anc
of the other factors and made difficulties in clear identification high resolution images: examples with Quickbird, Ikonos that
of possible complex relationships. and EROS. Archives of ISPRS Symposium, Comm. IV, buy
For EROS, the results observed were at the limit of the Ottawa. Ontario, Canada, July 8-12, 2002, Vol. 43, Part 4, 200
specification in case of basic scene and flat areas. The vector 547-551. son;
scenes gave hicher RMSE values, however, the ancillary data ; i. ; : mea
8 e A : E y D Toutin Th.: 2002. Error tracking in Ikonos geometric processing defi
sets cannot be considered as optimal (low accuracy GCPs) die : > > es tone
DA e aL ; : ^ using a 3D parametric model. Photogrammetric Engineering shox
making it is difficult to formulate reasonable conclusions for & Re ic Senung Guily9 = f
en e Sensing, 9,
these cases. Nevertheless, the tests showed that for “vector mote Sess, o mm
scenes" at least the twice as many GCPs are needed for ImageSat International. 2004. Products. ic
orthorectification process, compared to a basic scene. http;//www.imagesatintl.com/productsservices/proddetail/pr proc
The VHR validation program was a good opportunity to oducts.shtml (accessed 18 March 2004) prod
make a test and validation in more operational mode. It brought nnn ic Accus If th
. ^ . porn * 771 dani : cc ~ 3 e A ^Cu ~
a wide range of different experiences and results and can be Gradest T. Dial S: 2002. IKONOS Mp affe
considered as important contribution to the ^ process of Validation. Proceedings of ISPRS Comision 5 we conf
implementation of VHR satellite images in CAP oriented broad Symposium, Denver, CO, November 10-15, 200 and
the
area applications.
Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
ACKNOWLEDGEMENTS
Due to the limited space available, we are only able to
collectively thank all the contractors, image providers and
administrations who took part in this testing program.
1024
