International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004

Figure 1. The study-area of Thessaloniki, Greece

3. ORTHORECTIFICATION

As it is mentioned above, through the process of

orthorectification the two Spot images, the panchromatic and

the multispectral, are transformed into an orthogonal projection,

which makes them as accurate as a map in the same scale. This

is accomplished through the following procedures.

3.1 Geometric model and projection

The accurate geometric model, taking into account the

geometry of the acquisition and by recovering the distortions

that exist describes the relation between the sensor and one

ground reference system. In classic Photogrammetry this

relation between camera, image and ground is described by the

collinearity equations. This cannot be applied in remote sensing

because the elements of the exterior orientation are not stable.

This is due to the great speed of the satellite and the long time

of acquisition.

In the case of the spot geometric model, it is considered that

the satellite has steady movement during the time of the 9

seconds (duration of the acquisition of one spot image) and one

scanning line is defined as a reference line. This can be the first

or the central line. In this way, and as the exterior orientation of

the reference line has been defined; the exterior orientation of

all the rest lines is also determined, based on the distance

between them, the changes of the position of the project center

and the rotation angles.

The mathematical spot model is a broaden model of collinearity

equations with a difference that the satellite's route is steady

during the acquisition time so its movement can be modelled

with a polynomial of 1* and 2" grade. The exterior orientation

in a polynomial of 1* grade includes 6 parameters that describes

the route (position and speed),

X (=X, + X't (1)

YyX,- Yt (2)

Z(09Z,* Zt (3)

where — Xo, Yo, Zo * coordinates of projection center of

reference line

X’, Y’, Z’ = velocities in each direction

and 6 parameters that describe the satellite orientation,

ot)» o, *o't (4)

140

e(t)» Qo +t (5)

K(t)- K, -k't (6)

where ®, Q, K,= 3 rotation angles

Q0, Q9, K'7 their change rates

The result is a model that has 12 unknown parameters and the

time as an independent variable for the 6000 lines of a spot

image.

The Erdas IMAGINE has the option of choosing this geometric

model during the process of the orthorectification.

As far as the projection is concerned, the EGSA'87 is chosen,

because this one is the official projection of Greece.

3.2 DTM (Digital Terrain Model)

The integration of DTM into the two images is the most crucial

part of the orthorectification process, as its role is to eliminate

the relief displacement. Moreover the DTM quality affects the

accuracy of planimetry in orthophotos. The DTM that is used is

a mosaic of two different DTMs in order to succeed the utmost

accuracy. Both of them were produced with photogrammetric

proceedings from aerial photos of scale 1:10000 and

1:20000.The grid size is 25 meters.

3.3 Control points

The role of the ground control points (GCPs) is to define the

exterior orientation elements of one image. Their accuracy is of

a great importance because affects the accuracy of the

orthoimage. Except for their high accuracy, it is

recommendable that they have good geometry. This means that

they should have symmetric position over the image and cover

the whole area of interest.

At this point it should be mentioned that the ground control

points in this project were obtained from GPS measurements

and from other existing orthoimages that depict the same area.

For this reason they have different accuracy.

For the rectification of the panchromatic image, 57 control and

15 check points were used, and for the multispectral 61 control

and 15 check points.

3.4 Resampling

Most geometric transformations lead to pixels that do not

coincide with the original image. Resampling is the process of

calculating the intensity of the new pixels with one method of

interpolation. The data values for the pixels are interpolated on

the new grid from the values of the source pixels, an invaluable

procedure in the generation of an orthoimage.

In Erdas IMAGINE the following methods of interpolation are

supported:

|. Nearest neighbour: uses the value of the closest

pixel to assign to the output pixel value

2. Bilinear interpolation: uses the data file values of

four pixels in 2 x 2 window to calculate an output value

with a bilinear function

3. Bicubic interpolation: uses the data file values of

sixteen pixels in 4 x 4 window to calculate an output value

with a bicubic function