the ground
te pixel in a
5 paper the
of an object
ope € and
(€ =0), the
or 20/2 m.
adius R
(1)
posed that
the original
1d generally
itellite. The
ave left the
e may only
stead of all
otween the
f the grid p
| shift and a
st from o to
figure 3(2)).
oe resolved
he original
2 4(1)). The
en the two
is distance
-axis. Then
and Y =ly,l.
ect C, can
-grid, if its
(2)
n the
R should
T
p| +
p
|
].«3
:
Say (3)
2 2
It is known from the above calculations that the
demanded object radius R has no relationship with a turn
of the grids around o. Therefore only the grids' shift needs
to be considered.
2.2 Shift extent of grids
The lower bounds of the shift amount X and Y are
obviously zero (figure 2). What are the upper bounds?
See the figure 2 and the figure 5. The figure 2 shows the
original position of the grids. In the figure 5(1) X and Y
are smaller than a/2. In the figure 5(2) X and Y are
equal to a/2. Contrast the figure 5(1) and 5(2) with the
figure 2. It will be known that the demanded object radius
R is longest in the figure 5(2) and shortest in the figure 2.
In fact R increases with the increase of the shift's
distance, if 0O x X «a/2, Ox Y «a/2. How does R
change, if X 2 a/2, Y 2 a/2?
ij d Ixpl 0 < ly g
— xp SQ, O0 «lyol s À
d Pisa
The figure 6(1) shows such an example. In this situation
the neighbour 7, of the grid p is a nearest grid to the
center of the object C en C can be detected by a
complete SPOT-detector-grid, if its area is large enough
to include the grid 7. In fact it may be held that the grids
are formed by the method that a grid appears repeatedly
at intervals of 10m or 20m on the x-axis and on the y-axis.
The grids have the same periodic characteristic on the x-
axis and on the y-axis in other words. Therefore the
distance ofa may be treated as the equivalent shift of the
grid p. But it is tenable: O « Ixigl € a/2 and O « lysel x 8/2.
a a
Doce ael sit:
The figure 6(2) shows an example of this situation. It is
similar to 1) that the distance 0f¢ may be regarded as
the equivalent shift of the grid p. It is also tenable: 0 < Ixsl
€ a/2 and O « lysel x a/2.
a a
3) > Ixy € a, 5 « lypl € da x
See the figure 6(3). Similarly 0f; may be regarded now
as the equivalent shift of the grid p and it is tenable: 0 <
Ixzl € a/2 and 0 < lyezl < a/2.
4) Ixpl > A, Or lypl > A Or Ixpl > A and lypl > A:
See the figure 6(4). No matter how long the distance of
the shift is, a grid t' can be always found in the
neighbourhood of the object's center o, which meets 0 <
Ixel < a/2 and O< lyrl < a/2. ot’ is the equivalent shift of
the grid p.
Therefore the upper bounds of X and Y are equal to
a/2. The extent of the grids' shift is expressed as Os
X « a/2, 0 « Y « a/2. In this extent the demanded object
radius R increases with the increase of X and Y. R
should reach its maximum, when X - a/2 and Y= a/2
37
appear at the same time. The following is its
mathematical proving.
2.3 The sufficient area condition
Suppose Z-R^ (R50). Then the formula (2) is turned
into the following form:
Z2 yrs
= 2 Dur
Osxedcosrs2,. (4)
2 2
The region D, defined by 0x X x a/2, 0x Y « a/2 is
bounded and closed (figure 7). Therefore Z has a
minimum and a maximum on D,- The former is already
known (figure 2). The latter is the sufficient area condition
for an object on the smooth ground. Let us first consider
that to which direction the grid p shifts so that Z can get
a maximal increment. For this purpose the gradient of Z
at the point (0,0) is calculated: SZ. = (4,4).
The direction of KZ om is C. m, AS. Suppose
that the grid p shifts with a very small distance Ó in this
direction. The new coordinates of the grid p are
(X,, Y,) Certainly itis tenable: X, 2 Y. VZ at the
point (X, —Yj is calculated as
VZ|
equal to 45^. The direction of VZ does not change, if
the grid p shifts again by Ó in the direction 45°. If the
above process goes on, the gradient line originating from
O can be obtained. It is a straight line from the origin
O(0,0) to the point K(a/2, a/2) (figure 7). Therefore Z.
appears at the end of this vector X... — Y,,.- a/2
(figure 8):
follows:
(xx) ^ (2X, t a, 2Y; t a). Its direction is also
max
a ZZ ES dE (5)
553
is equal to d. The
sufficient area condition S, is calculated as follows
(figure 5(2)):
S, TRUE RU
Obviously the maximal radius R ax
(6)
For a panchromatic SPOT-image there are
d-1042 (m) and $,-200z (m.
For a multispectral SPOT-image there are
d=202 (m and $,-800z (m.
An object on the smooth ground can be detected at least
by a panchromatic (multispectral) SPOT-detector-grid, if
its area on the ground can include a circle with the radius
1042 m (2042 m). This radius is equal to the diagonal
length of a panchromatic (multispectral) SPOT-image
pixel.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B1. Vienna 1996