the topological structure generates. can not ‘perceive” the lines in a global im
Starting at each original interior point, context like we do, this procedure may Fi.
a search procedure is performed to left produce more critical points than that by CO
(or to right) until it meets with a manually collected.
boundary pixel. Then trace the boundary A
according to the right first rule (Fig. 11) 3.4.2 Geometrical rectification It is in
and mark the traced pixels. In the course of necessary to rectify the are points la
tracing, arc number is sorted in the arc because of the deformation caused by the t
file and filled in the polygn are file, orientation of the CCD camera. According me
and the polygn number is filled in the to the image orientation points in each la
items of left or right polygn number in the patch and their map coordinates, the qu
arc file, As for the arc whose start and end compressed vector data are rectified to à An
nodes are the same as another arc (Fig. 12), unique map coordinate system by affine Re
an arbitrary point on the arc is cooperated transformation. Because the algorithm ac
with the two rectifies the lines directly and avoids the
computation of background points, it A
y 4 operates fast. of
pr
9 3 3.4.3 Mosaicing Many processes are performed 0p
in patchs. The points on the border of the ie
2 window may deviate after each patch to be gl
j : ; : : rectified to the map. So mosaicing is -C
Fig.11 Right first Fig. 12 Arc identification necessary. A method like that in ( Beard re
rule 1986 ) is used in the system, First, the ex
: 3 nodes in the matching borders are extracted or
nodes to identify an arc number. When the If the distance of nodes in the two of
tracing stops at the end of a hanging up matching borders is little than a threshold, ma
line, a backforward tracing procedure is they are matched and their coordinates se
performed from the end point to the node, change to the same value in the arc and st
the arc number of the hanging up line is point file. Sé
filled in the polygn arc file and the pixel re
marks are deleted. After a polygn is 3.65 Attribute coding and geometric measuring th
enclosed, the odd and even test algorithm
lé. performed to fill the polygn. The The attribute codes of polygns and arcs REI
points in the polygn are marked and holes need to be input interactively. The areas [1]
are searched in the mean time. The data and lengthes are automatically measusured
structure of a hole generates in the same according to some standards. Various [2]
way and the polygn number of the hole is statistics can be easily caculated
filled in the hole file, according to the attributes. As a result,
; * ; statistical tables are printed, areas and
3.4 Vector data compression, rectification lengthes are automatically input to the [3]
and mosaicing thematic information base.
3.4.1 Vector data compression The vector data 4. EXPERIMENTS AND CONCLUSIONS
points gets from raster data are very dense, [4]
So it is necessary to compress the vector Many pieces of thematic map are
data, The system adopts the "local pixel processed by the system Fig.13a shows the
logic” method ( Hung 1988 ). The line following result of a kilometrer
compression procedure identifies critical grid. An enlarged view of line following is
points on an arc and keeps the node or shown in Fig 13b. Fig. 14 shows dashed
end point. Critical points are defined as line following. Fig. 16 shows arc [5
those feature points on a thin line such vectorization, Fig 16 shows the filling
that the line can be approximated by a result when generating the topological
connected set of straight line segments. data structure. (only paints green points
The algorithm traverses the chain code on the left and right sides of the polygn)
for each line and puts out a list of Fig. 17 shows the elarged view of the [6
critical points after a set of caculation
and identification. Since the computer
242