The QuickBird image products can be classifieds as into basic 
images, standard images, ortho-photos and stereo-photo pairs 
usually. They are also classed into panchromatic band image 
data, multispectral image data, products package of 
panchromatic band image data and multispectral image data and 
fusion image data (true color or false color) based on the 
integrated bands. 
2.2 The brief introductions of experimental satellite photos 
The experimental satellite photos are single panchromatic of 
standard products and multispectral satellite photos taken on 
January 7, 2003 of USA time in Fuxin City of Liaoning 
Province. The ground sampling resolution is 0.636m, the 
reference ellipsoid is WGS84. the projection is transverse 
Mercator projection, the heliacal horizontal angle is 159.671° 
and the heliacal zenith angle is 23.0836°, the satellite horizontal 
angle is 99.4251°, the satellite zenith angle is 76.867°, the 
orbital perspective angle is 0.0473414° and the vertical orbital 
perspective angle is 12.2236°. The satellite position and 
situation are measured accurately so that the quality in good 
condition. The ground intervals both in longitudinal and lateral 
directions are 0.6m and the mode is quadrate convolution for 
resample. The other information is almost the same of 
multispectral satellite photos and panchromatic satellite photos 
except that they all have red, green, and blue bands. Meanwhile, 
the intervals are 2.542m in ground sampling and 2.4m in 
resample. 
3. PRECISION EVALUATIONS 
3.1 Images preprocessing 
3.1.1 Clipping and fusion of images: The human eyes are 
limited in the ability of density of black and white, in which 
only can reach ten gray-levels, but is higher in the 
discrimination of color images. If the average resolution of 
human eyes is AA — 3uiYl , more than hundred of colors can 
be distinguished and this only is one of elements in colors. If 
the other two factors saturation and lightness are considered, the 
level of distinguishing different colors for human eyes is far 
greater than that of distinguishing between black and white. In 
order to take advantage of colors in remote sensing image 
interpretation, panchromatic images and multispectral images 
can be merged, so that the characteristics of both the high 
spatial resolution of panchromatic images and the interpretation 
of multispectral image can be reserved. The main-elements 
analysis method of Erdas is applied in fusion, the quadrate 
convolution mode used in resample and a combination of 
bands-1, -2 and -3 in image display. Because the range of 
satellite photos we got is very large about 200 km 2 , the amount 
of data is too much, that is, the panchromatic satellite photos 
are about 1.7 Gbit, the multispectral satellite photos are about 
700 Mbit, and the data after fusing is near 7 Gbit. Considering 
the data quantity and convenience in data processing, an image 
around the Liaoning Technology University approximately one 
square kilometers scope is captured where the relief is flat (the 
difference in elevation is less than one meter) and crowded by 
buildings. 
3.1.2 Determination of internal geometric precision: The 
QuickBird standard image is corrected by using a rough DEM 
model, and ground and land features are standardized on 
WGS84 reference ellipsoidal surface, so that it isn’t 
orthographic images, because the degree of precision is quite 
low. The QuickBird standard images should have better internal 
geometric precision. Hence after systematic correction, 
generally speaking, if the interior geometric error of satellite 
image is smaller than or equal to a pixel, it will satisfy the 
precision request in the most remote sensing application. Erdas 
remote sensing software carries on the image geometric 
rectification by the way of the integral correction, and if the 
internal geometric precision high, then the image precision will 
be better after rectification. Therefore, an analyzing to the 
internal geometric precision of the empirical data should be 
undertaken. The concreting method is that by choosing 15 
control points on the topographic diagram to discover the 
corresponding image points in the fusion image, the root mean 
square of length difference between each line connecting 
control points on the image and its corresponding lines on the 
topographic map, that is 
Where, Ad t = d mi - d gi , (z = 1,2,3, ••*,«), d mi is the 
length of the line i connecting the control points on the image, 
d . is the length of the corresponding lines on the topographic 
map, and n is the total number of the lines between control 
points. 
Fifteen control points are selected in this experiment, therefore 
there are 105 control sides altogether, and after 
calculation, 8 = 0.5723m < 0.6m it is smaller than a pixel 
in which indicates that the internal precision is quite high from 
using the QuickBird standard images. 
3.2 Accurate geometric rectification of images 
For the purpose of relationship registration of the remote 
sensing images and the geographic space on site, which is 
called location and orientation, the geometric rectification to the 
images must be carried out before the evaluation of the interior 
position precision. The conventional methods for the geometric 
rectification are divided into optical rectification and the digital 
one. At present, since the images are usually based on the 
digital data, the digital geometric rectification is generally used. 
Actually, there are two kinds of methods applied: one is the 
collinear equations using digital elevation model (DEM) and 
corresponding imaging equation and another is control point 
computation method based on a certain mathematical model, in 
which the polynomial method is typical one. Theoretically, the 
method of collinearly equations is stricter than the polynomials, 
specially introduced the ground elevation information in the 
process, so that is superior to the polynomials in the case of 
ground undulation. Although the method is higher in precision, 
but the DEM model must be established firstly, otherwise the 
rectification will become difficulty and complicated if lack of 
the digital ground elevation. Moreover, in satellite dynamic 
sensor, the position and the attitude angles of sensor are