database, ie. Um is the universe of M. We will assume that the
geometry of the objects is represented in a vector format with a
full topological structure, i.e. the geometry is described in nodes,
edges and faces defining a geometric partition of the mapped area
(or O-. 1- and 2-cells). Let Geom(M) be the geometric component
of M, ie. it is the collection of all geometric elements describing
the geometry of all objects of the universe. Let Face(M) be the
collection of all faces in Geom(M); similarly Edge(M) is the
collection of all edges and Node(M) is the collection of all nodes.
The function Part, [ f. O] will be introduced to express the
relation between a face f € Face(M) and an object O € Uy. If this
function has the value = 7 then the face belongs to the spatial
extent of the object, if the value = 0 then that is not the case. We
can now define the set:
Face(O) = { [1 Party: [ f, O] = 1}
Face (O) is the spatial extent of O. In this notation the geometric
description of the objects is organised per object. For each edge e
we can express its relationship to a face f by the functions:
Le[ e, f] 7 l if e has fat its left-hand side and
Le[ e, f] - O otherwise,
and similarly
Rif e. f] = I if it has fat its right hand side and
Ril e, [| = 0 otherwise.
With these functions the relationship between an edge ¢ and an
object O can be established:
Le| e, Ol f] 2 MIN( Le e. f], Party] f. Of) and
Ril e, Ol f] 2 MIN Ril e, f], Partzsl f, O])
Object O Te
ut / \
ee A A
IT gy
fp
PTT
à Part,[f, O] 2 1
I
Figure 1: Rclationship between edge, face and Object
If the edge has the face at its left hand side and the face is part of
the spatial extent of the object then both functions on the right
hand side of the first expression have the value = / and therefore
the first expression gets the value = /. This means that the edge
has the object at its left-hand side and thus we get Le/ e, O | I4=
/, otherwise it will be = 0. Similarly if the edge has the object at
its right hand side then Ri/ e, O | f] = 1 otherwise = 0. If there is
a face for which Le[ e, O | f] 5 1 then that implies that the edge
e has object O at its left-hand side, so that Lef e, 0} =l
Otherwise Le[ e, O] 2 0. and similarly for Rif e. O |. With these
two functions we define:
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV. Part B4. Istanbul 2004
Ble. Oj=lefe,O] +Ri[e, Of
If the edge has the object at its left and right-hand side this function
has the value 2. If the object is only at one side, so that the edge
belongs to the boundary of the object, the value will be I. If the
edge is not related to the object then the value will be 0. The
boundary of the object is therefore the set of edges for which this
function has the value 1:
90-[elB[ e.O] 2 1]
3. SPATIAL AND THEMATIC PARTITIONS
The universe U,, is a complete coverage if for every member f of
Face(M) there is at least one object O so that Parts, [ f, O] = 1.
That means that objects cover the whole area covered by the
geometry of M.
The universe is a spatial partition if it is a complete coverage and
if the objects do not overlap. i.e. for each face f of Face(M) there
is exactly onc object O so that Part»; [ f. O] — I.
| Nat orsid | | forest I agricult |
Figure 2: A collection of classes forming a thematic partition
Now let P={ CI. C2. ... ,Cn} be a collection of classes so that for
each i we have Ci © Um. P is a thematic partition of Uy if each
object is a member of exactly one class. This means that the
classes are properly specified so that the thematic description of
the objects is unambiguous, sce Figure 2 and 3.
If P is a thematic partition of Uy, and Uy, is a spatial partition of M
then P. generates a. spatial partition of M. That means that the
classes cover the whole mapped arca and they are spatially
distinct; see (Molenaar 1998).
Faces Objects Classes
2 = Object 1 ay,
7 os Object 2 = [1 Natgrssind
9,10,11.... "Object 3 ayy +}
3,4, == Object 4 «2 Forest
1; 5,6, 8, 12== Object 5 = Agriculture
Figure 3: Objects and classes Form a dual partition structure, je.
the combination of a spatial and a thematic partition
Interne
Tt
Th
Ey
Th
U,
Th
P:
We
ob
be
We
pat
are
pai
Hi
If «
twe
one
ex
The
ord
leve
hiei
eac
eac
The
(will
class
prop
ever
stric
next
class