nbul 2004 
ensing 52 
. Digital 
13-762 
tonomous 
rammetry 
n 
ROBUST DETECTION OF ROAD JUNCTIONS IN VHR IMAGES USING AN 
IMPROVED RIDGE DETECTOR 
S.Gautama, W.Goeman, J.D'Haeyer 
Dept. Telecommunication and Information Processing, Ghent University 
St.Pietersnieuwstraat 41, B-9000 Gent, Belgium 
sidharta.gautama@ugent.be 
KEY WORDS: GIS, object recognition, road detection, change detection 
ABSTRACT: 
In this paper we present a novel technique for automatic detection of road junctions in VHR satellite images. The detection is based 
on an initially detected road network using a differential ridge detector. This is a standard technique for road detection. However, the 
performance of .the ridge detector degrades in the vicinity of junctions because the line model on which it is based does not hold 
anymore. We analyze the content and quality of the derived information layers of the ridge detector and show which information is 
useful for the detection of junctions. The detected network is improved using a region growing and thinning strategy. Junctions are 
detected based on this network using a shape analysis. This analysis puts restrictions on the appearance of junctions and allows for 
an efficient filtering of false alarms. Experimental results performed on IKONOS images over the city of Ghent show a reasonable 
detection rate with a very low false alarm rate can be achieved. This low false alarm rate is important for our purpose of quality 
assessment as it requires reliable image information to make a robust comparison with the road database. 
1. INTRODUCTION 
A major challenge in the production and use of geographic 
information is assessment and control of the quality of the 
spatial data. The rapid growing number of sources of geospatial 
data, ranging from high-resolution satellite and airborne 
sensors, GPS, and derivative geospatial products, pose severe 
problems for integrating data. Content providers face the 
problem of continuously ensuring that the information they 
produce is reliable, accurate and up-to-date. Integrity 
constraints are able to resolve certain issues in the data, like 
valid attribute values or relationships between data objects. The 
main issue is however the consistency of the data with respect 
to the current "real-world" situation. Today the industry still 
relies on human operators, who collect and interpret aerial 
photos and field data to check and correct the current state of 
the data. Especially for detailed data over large regions, like 
digital road maps or topographic maps, this is a very labour- 
intensive and costly process. In addition, human processing is a 
source of error and inconsistency. Automated detection of 
change and anomalies in the existing databases using image 
information can form an essential tool to support quality control 
and maintenance of spatial information. 
The main problem however are the differences in data 
representation. To be able to compare geospatial vector data 
with images, the information in the images needs to be 
described in terms of object features. Automatic detection of 
man-made objects is a difficult problem. Shadow, occlusion and 
variety in appearance all give rise to a fragmented and 
imprecise description of the image content, especially if 
consistent detection is required over large datasets. Within the 
field of automatic quality assessment, there is a high need for 
powerful detection techniques but with a strong emphasis on 
reliability. From the viewpoint of the data provider, a statement 
about the quality of his data is only useful if the statement can 
be made with high reliability. 
In our work, we investigate a system for change detection based 
on object based spatial registration, where detected object 
features in the image are registered to corresponding features in 
815 
the vector data. The system consists out of two stages: 1) a low- 
level feature detection process, which extracts information 
about the road network, and 2) a high-level matching process, 
which uses graph matching to find correspondences between 
the detected image information and the road vector data 
(cfr.Figure 1). The graph matching process is driven by the 
spatial relations between the features and takes into account 
different errors that can occur (e.g. spatial inaccuracy, data 
inconsistencies between image and vector data). The matched 
features can be used to calculate a local transformation between 
image and vector data, which is able to compensate for local 
distortions that can occur between the datasets. Additionally the 
object-to-object mapping is useful to define measures of change 
between datasets. 
         
Ld T € i 
disc E "s 
spatial registration EX]. 
and EE 
change detection 
  
report containing 
measures of change 
  
Figure 1. System overview for object based quality assessment. 
Although much effort has been spent on designing algorithms 
for road detection, the complexity of the problem is still not 
fully tackled. In this paper we examine road detection using a 
ridge detector. Lines in an image can be seen as narrow valleys 
or ridges in the intensity surface if one views the image as a 
terrain model. This ridge model works well for roads which are 
bordered by homogeneous regions like fields. However in the