ibllshed by the
mtour-line extrac-
case when a part
isually on photos*
boundaries of plots
i natural borders,
rtraents differing
st mensuration
ity, age, etc*,
sj relief, lines
isotherms, etc*
st management map-
os* For this pur-
ering of contour
ne has been deve
rst transferred
photograph in whi-
e up a photo out-
from the photo-
ered into the sys-
0 outline conto-
xtraction is done
kage which reali-
sharp' and *logi-
ethod the positive
e between sharp
ned* It contains
about contour li-
e multiplied by
to the source
on the display
d by means of cha-
1 nonsharp' mask
, and contour line
rther processed
Ls a scanning win-
nparison of brigh-
and pixels of the
si P and with each
le set logical
sting the conditi-
narked which cor-
of the contour
*
: the contour net-
ixtraction. It is
>f the residual
:hickening, thin-
, etc* The proces-
>f the analysis of
>ods• The neighbo-
:ive of the conto-
itral point, that
:ision about the
j in the central
:ects class defi-
Their neighbour-
latter marked or
iple, to eliminate
>f cells in which
¡ent are analysed*
ige cells in which
led the ’current
t quite broad nei-
lell M . We shall
tg-shaped zones
etc* The side of
■1 elements of the
elements in the
s to (2n+l)4-4»
its make up one
shown in a code
corresponding to the n-byte word*
Let us designate the first zone of 3x3
around the cell M through Z,.* If we cut and
develop it we shall have an octal-digit by
te the contents of which reflect the state
of the neighbourhood M . Let's divide this
byte into X-half-byte ind Y-half-byte* By
marking the contents on the XOY system da
tum lines we shall get the matrix of 16x16,
each element of which corresponds to one of
the Z*-neighbourhood states*
By means of grouping the states of the
neighbourhood Z* by the definite feature
into classes one can extract areas of these
classes from the matrix and mark them with
classes names: A, B, C, etc* For example,
class A shows breaks of type I-I with which
ends of the broken circuit get into the nei
ghbourhood of 3x3. A more detailed descrip
tion of different classes of the Z 1 -neigh
bourhood states is given in (Elman 1 1982).
Each class must correspond to the solution
rule of the type: 'if A, then the unit must
be sent to M ', or 'if A, then zero must be
sent to M '*
Each restoration procedure includes the
following main operations: the current cell
search, analysis of Z--neighbourhood of ea
ch current cell, formation of the code-state
and determination of the contents of X- and
Y-half-bytes, entrance along X and Y into
the state matrix to pick up the defect class
and fulfilment of the decisive rule and tr
ansition to a new cell. In this way we ma
nage to single out up to 13 classes of con
tour defects*
3 CONTOUR NETWORK PROCESSING
The binary contour network restored in the
system is to be further processed so that
it can be reduced to the form necessary for
the submersion into the cartographic data
base and for plotting* Certain demands are
made to the network contour lines: all the
lines must have standard thickness of one
pixel of the display image, contiguous pi
xels of the line must touch each other with
their sides, there must be no breaks in
closed circuits, etc*
During the processing the transformation
of the contour image from the pattern into
the vector form with the cutting of the who
le ramified network into separate unramifi
ed sections called 'threads , the coding of
threads depending on the meaning contents
and position in the network, the analytical
rectification of all contour lines hy the
revision points read from photographs and
the map, and the network map control are
carried out*
In order to process complex contour net
works the digital sensing method has been
developed (Elman 1982), On its basis the
so called 'transport' algorithm is built
which allows to follow any contour network
along the lines in the forward and opposite
direction* In each point of the contour the
necessary processing can be done*
The essence of the method lies in the fol
lowing* Above the binary image located in
the marker memory of the display system the
inquiry program-control element (the digi
tal probe) is defined* The probe can be mo
ved by the program in the binary memory from
the element of the digital image to the ad
jacent element* Let relocations only in fo
ur directions be allowed: to the right, to
the left, up and down, and not allowed along
the diagonals* We shall number this direc
tions with figures from 0 to 3* When the
probe from the current call moves in each
of these directions we shall first call the
se figures the direct addresses, and when
it returns - the return addresses*
We shall bind two more matrices with the
same addressing with the matrix of the bi
nary memory* They are intended for storing
the direct and back addresses of the next
probe step* When the probe makes the next
step in the forward direction from the cur
rent cell, the next address of the future
direct step is entered in the memory of di
rect addresses of this cell (before its mo
vement begins the direct addresses matrix
is coded with the initial address, for «ca
mple, zero)* When the probe enters a new
cell, into the memory of the return of this
cell the address of the return from it into
the former current cell is put* From each
contour cell all necessary direct steps must
be made in succession, and after the opera
tion is over the return step has to be made*
The network inspection is done in the fol
lowing way* First the probe is put into the
arbitrary contour network point, for exam
ple by means of scanning* This point is con
sidered initial, and is marked* For this
point the initial direct address is extracted
and by this address the neighbourhood cell
sensing ls accomplished the aim of which is
to test the three conditions allowing the
movement of the probe into this cell: if
there is a contour (marker) in this cell,
if there was the probe there aarlier, and
if the probe went beyond the frame borders*
In meeting all these conditions the direct
step into the sensed cell is made* Otherwise
the next direct address is extracted* If all
four direct addresses are exhausted then it
is tested if the given cell is initial, and
if so the network inspection is completed*
Otherwise the return address is extracted,
and the step to this address is made* In
such a way all the contour network is ins
pected* In the end points the probe is pre
sent once, in linear points - two times
(in the forward and opposite movement), in
focal points - three or four times depen
ding on the number of branches, and in the
initial point - one unit more than in the
point of the corresponding type.
It must be noted that the probe can go
through all the points of the arbitrary con
tour network, even in case when the network
has the broadening in the form of areas
flood by the marker and when there are ho
les in the areas*
The described method is also used for the
analysis of individual areas of the arbit
rary form with holes, and not only contour
networks* The processing of the contour
olnts or area can be done after each fifth
orward step (A-processing), before and af
ter the return step (B- and C-processing),
In order to transfer the network from the
pattern into the vector form the following
operations are carried out: with the help
of digital sensing each closed area of the
arbitrary form is described with its number,
the probe's movement around the contour net
work is set, on the direct step all focal
ints are revealed and marked, and on the
ck step chains of the threads coordinates
are formed each of which begins and ends in
focal or end points* To each thread numbers
of two areas lying on its sides are assign
ed* As a rule the thread is represented in
the machine as a broken line that is why
