Upload
hafwen
View
37
Download
1
Embed Size (px)
DESCRIPTION
ICA workshop on geospatial analysis and modeling. Implementation of a generic road-matching approach for the integration of postal data. M. Zhang, L. Meng. Department of Cartography, TU Munich. Motivation The matching strategy Matching results Assessment of matching certainty - PowerPoint PPT Presentation
Citation preview
1
Implementation of a generic road-matching approach for the integration of postal data
M. Zhang, L. Meng
Department of Cartography, TU Munich
1. Motivation
2. The matching strategy
3. Matching results
4. Assessment of matching certainty
5. Application – Integration of postal data
6. Conclusion and future work
ICA workshop on geospatial analysis and modeling
Vienna, 2006-07-08
2
Task Enriching Basis DLM with post addresses, which are
bound to the road layers of TeleAtlas.
1. Motivation - Project “Postal data integration”
Problem Absence of some important road attributes, such as street name, etc.
Solution Matching between Basis DLM und TeleAtlas
3
2. The matching strategy
Data Matching
(1) Data preprocessing
(2) Matching of road objects
(3) Unsymmetrical Buffer Growing
(4) Conflict solution
An integrative matching process
Data integration
4
2.1 Data preprocessing
(a) Reduction of noisy or irrelevant details
5
2.1 Data preprocessing
Example
TopoR_Node = 1 2 3 4 5 6
AA
1
0
Classification
A
4TopoRTyp
(b) Description of topology with node geometry
The direction of AA‘.2
The direction of AA‘.1
0
Classification
AAa
ß
r
A
A’
3TopoRTyp ,3TopoRWinkel )2,0[
.
A
A’
6
2.1 Data preprocessing
(c) Reduction of topological differences
A0
B1B2
B0
Point B 1 : B1 B0
Point B 2 : B 2 B0
Line B 1B2 : Delete
: objects in dataset 1
: objects in dataset 2
: improved objects in dataset 2
Improvement process:
7
2.2 Matching of road objects
Step1: Instantiation of the reference polyline
8
2.2 Matching of road objects
Step 2: Identification of possible matching candidates
Step 3: Exclusion of incorrect candidates
α s
A
G
O
O’
Matching candidates
Matching reference
Turning points&Buffer-P
Buffer-V
H
K
9
2.2 Matching of road objects
Step 4: Exactness inspection of the matching candidates
A
G
H
K
O
O’
Matching candidates
Matching reference
Turning points&Buffer-P
Buffer-V
)_,_(_ SimilarityTopoSimilarityGeoFSumWMatching
]1,0[_ SumWMatching0: an entirely wrong match
1: a perfect match. {
10
2.2 Matching of road objects
Step 5: Establishment of link file
11
2.3 Unsymmetrical Buffer Growing
46 M
41 M
A
B
C
D
The mean and variance of the position discrepancies between the matched pairs can be used as the indicators to capture the global deviation in the current matching area.
12
2.4 Conflict solution
Conflict
SolutionCriterion: Larger or Largest
“Matching_SumW”
13
3. Matching result
Area (a) Area (b) Area (c) All
Number of line objects
Basis DLM 311 670 362 1343
TeleAtlas 319 733 459 1511
Number of points along road lines of TeleAtlas
804 2413 1611 4828
Number of matched points
Correct match 766 (95.3%) 2292 (95.0%)
1583 (98.3)
4641 (96.1%)
Poor match 8 (1.0%) 16 (0.7%) 16 (1.0%) 40 (0.8%)
Number of points with no match 30 (3.7%) 105(4.3%) 12 (0.7%) 147 (3.1%)
Computing time 26 seconds 65 seconds 42 seconds 133 seconds
(a) (b) (c)
14
4. Assessment of matching certainty
(a) Definition of the Variable „Matching_Certainty“
Definition {IF Conflicting match THEN
Matching_Certainty =ELSE
Matching_Certainty =END IF}
? ?
Matching Errors
}20,0;_{ SumWMatchingMin
SumWMatching _
15
4. Assessment of matching certainty
(b) Distribution of “Matching_Certainty”
0
2
4
6
8
10
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
Matching_Certainty
Den
sity
falsely matched objects
overall matched objects
Matching_Certainty [0.00, 0.20] (0.20, 0.75) [0.75, 1.00]
Density distribution
Among the falsely matched objects ca. 90% ca. 10% 0 %
Among the overall matched objects ca. 25% ca. 50% ca. 25%
16
4. Assessment of matching certainty
(c) Classification of the Matching Certainty
Class of the matching certainty
perfect good possible
Matching_Certainty [0.75, 1.00] (0.20, 0.75) [0.00, 0.20]
17
5. Application – Integration of postal data
Integration process of postal data
18
6. Conclusion and future work
Future work Refining the current matching approach
Dealing with the special matching cases
Conclusion Unsymmetrical Buffer Growing - generic nature
Definition of Matching Certainty and classification of matching results - a more comfortable interaction.