ISPRS, Vol.34, Part 2W2, ‘‘Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001
241
Road Sign Recognition is a field of applied computer vision
research concerned with the automatic detection and
classification of traffic signs acquired from a moving car. Many
researches have been developed since 1984 in Japan but almost
Road sign recognition systems exist in developed countries.
Complicated, expensive systems such as feature extraction from
occluded images based in colors (Kentarnavaz, Estevez, 1994,
USA), color based signs finder for highways (Frank, Haag, Lee
Gang, Taiwan) etc have been developed but a simple, cheap
algorithm that is suitable for developing countries has not been
developed yet.
In Thailand, there is no research on road sign recognition
although “road accident’s account for the road is the highest
deaths of people annually” (National Statistical Office, 1998).
Moreover the development of highways systems changes rapidly
in each year. This reason makes the road database and road
map fast out-of-date. Thus the Mobile Mapping System and
Road Sign Recognition System that are not complicated, and
low-cost should be considered.
In AIT, some parts of the mobile mapping have been developed.
Improvement of the accuracy for RTK-GPS (Dinesh, 1998) and
Measurement the center line of road (Ganesh, 1999) are the
development of the position sensor, one part of mobile mapping
but another part, road sign image recognition, that is important
equally, has not been done.
Therefore, the purpose of this study is to develop the new simple
algorithm of the Thai Road Sign Recognition and Road Sing
Positioning. The algorithm is descried in the next section.
2 Sign Recognition
For this section, the study was separated to two steps: Sign
detection and Sign identification.
Sign Detection
The first step in the recognition of the sign is to identify the region
of Interest or sign shape. This step is the semi-automated
detection applied by affine geometric correction below:
The affine transformation is applied to solve both rotated and
scaling distortion of image due to the 3Dimension to 2Dimension
projection of a camera by formulae below
x' = ax + by
y' = dx + ey -
(1)
Figure 3 the procedure of transformation
Where
x, y is the coordinate of old system.
x\ y' is the coordinate of new system.
a, b, c, d, e, f are the parameter solved by using the least square
method below:
B = Ax -> x = (A T .A)' 1 .A T .B (2)
x'i~
*1
.Vi
1
0
0
0'
y[
0
0
0
*1
yl
1
x' 2
*2
y 2
1
0
0
0
y r 2
A =
0
0
0
*2
y 2
1
*3
*3
1
0
0
0
y'l
0
0
0
*3
-V 3
1
x 'r
[
*4
J4
1
0
0
0
J 4_
0
0
0
*4
y 4
1
