jse and 
re than 
section 
or. The 
ve. 
ode for 
espond 
101 ,::3. 
010110, 
1.3114, 
1. 20. 
ctly be 
; face- 
ver, a 
two or 
cessary 
-aspect. 
osition, 
'rgence 
of two 
y these 
'rgence 
de, the 
of two 
in base 
'rgence 
es are 
ids for 
inalogy 
5660666 
  
For mergences more than two face-aspects, 
the principle is that if only a code differs from 
others in base of face-aspect, these visual faces 
are projection of a 3-D primitive. For example, 
Fig. 5a is a rhombus, its mergence code is 1222, 
and only a code is 1, others all are 2. Fig. 5b is 
its aspect-graph. 
2.3 Coding Split Principle 
Some primitive-aspects can correctly be 
interpreted by face-aspect code and face-aspect 
mergence code. However, projection in fact is 
very complex. In the process of two primitives 
merging by Boolean operator, the face could 
sometimes be smooth, and its 2-D projection 
seems like a face after image segment. But it is 
in fact mergence of several volume-primitives. 
In order to successfully interpret aspect like 
these easy confusion, the scheme of split face- 
aspect is proposed. We conclude some 
projection situations, and find that when 
neighbor edges within the same face form an 
edge-tuple, if the edge-tuple has the same code 
as well as differ from other neighbor edge code, 
the aspect is mergence projection of two 3-D 
primitives. In this case, we must split the 
original aspect into two aspects. lts split 
principle is to connect the head and end with 
neighbor edge code (1 or 0). 1 stands for straight 
line, and 0 stands for curve. Therefore, the 
original code must be rearranged into two split 
aspects. For example, Fig. 6a shows a projection 
of both cone and cylinder, if the aspect 4 be not 
split into two aspects, its code is 510011. In 
term to split principle above, we must split the 
aspect À into aspects 1,3. The aspect | and 
aspect 3 are connected with dotted straight 
line[Fig. 6b] since neighbor code is 1(for 
straight). Their codes after split are 4111111, 
3100 or 3001 or 3010 respectively. 
  
(a) (b) 
Fig. 6 Face-aspect split 
So far, we can recognize some simple and 
regular primitives, which lost less information in 
projection, using aspect-interpretation including 
face-aspect, face-aspect mergence and face- 
aspect split. Nevertheless, most of primitives are 
still difficult to recognize correctly because 
original primitive-aspect is damaged by Boolean 
operator. So we still have to make full use the 
prior CAD information to achieve our aim. 
1 
International Archives of Photogrammetry and 
3. IMAGE ATTRIBUTE RELATIONAL 
GRAPH-ASPECT GRAPH 
There are a lot of different applications for 
CAD prior knowledge. We here use the CAD 
data structure to construct model attribute 
relational graph, and then implement model 
matching. Some definitions are in advance given 
as follows. 
definition 1 An attributed pair is an ordinal 
pair(An, Vd), where An is the property name 
described object, and Vd is the property value. 
For example, describing a face-aspect is S=(area, 
30), where area is property name, and 30 is 
property value. 
definition 2 An attributed set is an m-tuple [p!, 
p2, ... pm]. Where each element in the tuple is 
an attributed pair. For example, describing a red 
triangle is S: [(color, red), (type, triangle), (area, 
30)]. 
definition 3 An attributed relational graph of an 
object is a graph, represented G=(V, A), where 
V is a set of attribute node, A is a set of branch, 
sometimes called acre, used as describing 
connecting relation of two nodes. 
definition 4 A primitive attribute Graph is graph 
represented volume-primitive projected of 3-D 
primitive. 
definition 5 An attributed hypergraph consists of 
a set of hypernodes and hyperarce. Each 
hypernode stands for a primitive attribute graph. 
Each hyperarce stands for the relation between a 
pair of hypernode. 
  
(a) I F4 db) 
Fig. 7 Object and aspect-graph 
In term to definition above, it is easy to 
construct the attribute graph, primitive graph 
and hypergraph[Fig. 7b] for some simple object 
such as Fig.7a. But it is sometimes relatively 
difficult to construct the attribute hypergraph 
such as Fig. 8b since we not know which face- 
aspect or which several face-aspects mergence 
can represent a whole primitive-aspect. 
Nevertheless, with the help of  aspect- 
interpretation, it become very easy to construct 
the attribute graph, primitive graph and attribute 
hypergraph. So we now study how to construct 
the aspect-graph in term to aspect-interpretation. 
021 
Remote Sensing. Vol. XXXI, Part B3. Vienna 1996