The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
Thus the first step is the core of the updating process, and the 
key is to develop the algorithms for matching and selective 
3. MATCHING BASED ON ANALYSIS OF LEVELS 
Beside application in GIS data updating, network matching also 
plays an important role in image processing, data integration 
and so on, so considerable efforts have been devoted for 
network matching. These methods mostly focus on an 
investigation of algorithm that consists of three types of 
matching, i.e. node matching, segment matching, and edge 
matching. The matching measures for nodes matching include 
distance, the degree of node and the value of spider function 
(Rosen and Saalfeld, 1985; Saalfeld, 1988). Edge matching can 
be implemented based on the node and segment matching 
(Gabay and Doytsher, 1994; Filin and Doytsher, 1999; Lemarie 
and Raynal, 1996), or directly using of the measures and the 
buffer tool (Walter and Fritsch 1999, Badard 1998, zhang et al., 
2005). However, these approaches mentioned above are focused 
on matching of data with same or similar scales. 
3.1 Analyses for matching levels 
Indeed, a road network can be regarded as the composing of 
line feature and crossroad feature, so the analyses for matching 
levels are also divided into two parts according to the matching 
of two type features. 
One edge can be divided into segments and several edges can 
be joined a route. So for line feature, three levels are displayed, 
and we name segment as the decomposed level object, edge as 
the basic level objects and route as the abstract level object. As 
to line feature matching, three matching levels are also formed 
according to line feature levels (see figure 1). 
Abstracted level 
Route mathing 
, 
L 
Basic level 
Edge matching 
; 
к 
Decomposed level 
Segment matching 
Figure 1. Three matching levels for linear feature 
It is difficult for using measures of matching to realize edge 
matching of many to many. However, for matching of segment 
or route, the types of matching are either 1:1 or 1:0, due to 
segments or routes as the results of edges spited or joined. So 
segment or route matching can be easily confirmed using 
measures, and edge matching can be obtained based on these 
two kind matching. Moreover using of segment matching is 
frequently adopted methods (Xiong and Sperling, 2004). 
Depending on the criteria of length and direction segments can 
be split along edges with same scales. Nevertheless, for 
different scales segments split and their counterparts are not 
enough equal due to the different abstract levels, and it will 
affect accuracy of edge matching. It is not feasible using 
segment matching to obtain edge matching under this condition. 
omission. Then they are detailed in the next two sections 
A joined route and its counterpart can be easily built, so route 
matching is used to represent edge matching in this paper. 
Route matching is divided into 1:N and M:N (M>0, N>0) 
matching according to the number of edges on the routes 
corresponding (see figure 2), where 1 (or M) represents number 
of edges on the route at large scale and N represents number of 
edges on the other route at small scale. Edge matching can be 
regarded as the route mapping of one to one, and be induced to 
route matching. 
(a) Route matching of 1 :N (b) Route matching of M:N 
Figure 2. Two types of route matching 
Simple crossroads are all described as one node on two road 
networks at different scales. However, complex crossroads are 
represented differently on two road networks at different scales. 
These crossroads are composed of nodes and edges. So 
matching of nodes and matching between node and edge are 
defined. In addition, an end point can be regarded as the 
decomposed level of a node, and the matching end points will 
help to build node matching and route matching. 
Based on the analysis above, the crossroads corresponding can 
also be divided into three levels of matching. The end point 
matching can be regarded as the decomposed matching level, 
and the node matching as the basic matching level, and the 
route matching and matching between node and edge as the 
abstracted matching level, see figure 3. 
Figure 3. Three matching levels for crossroad feature 
3.2 Strategy and methods for matching 
According to the principle from the simple to the complex, the 
method of matching adopts the bottom-up strategy. It is starting 
with end point matching, then proceeding node matching, and 
finally ending up route matching and matching between node 
and edge. This strategy is consistent with the order of matching 
levels from the decomposed to the abstract. To obtain the higher 
levels of matching, they need to be transformed into the lower 
levels of matching, or using the lower levels of matching. For 
instance, M:N route matching can be transformed 1:N route 
matching and 1:N route matching can be transformed into 1:1 
route matching again by joining edges with different algorithms. 
In addition, the low level of matching may also need the help 
from the high level of matching. For example, the part of 
292