ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001
Where
A = a metric of the old coordinate
B = a metric of the old coordinate
x metric of the parameter
Because of a large number of the Euclidean Distance (E (i,j)), the
Euclidean Distance is normalized by dividing with some constant.
Here the power of image size is applied by formula below:
This step was the Sign extraction form the environment and the
geometric correction of the distorted image as size as the sign
image in Sign Library.
Sign Identification
The key part of this study is the identification of Road Sign type.
The research was simplified with the gray-scale image
processing. The image matching with Normalized Euclidean
Distance was considered in this step below:
NED
E(i,j)
(imagesize ) 2
Where
(5)
NED is the Normalized Euclidean Distance.
The image size = (a range of gray scale)*(the wide of image)*(the
length of image)*(number of color channel)
Euclidean Distance, simple Image Matching was applied for
Road Sign Recognition by formulae below:
E (*» j) = Z X 18 (*» j) - t(i, /)]
• j
Where
(3)
Compared with all sign image in Sign Library, The minimum
value of the Normalized Euclidean Distance with each sign
image in Sign Library identifies the type of the sign from a target
image.
3 Sign Positioning
E (i,j) is the Euclidean Distance between target image and
Template.
g(i,j) is a gray scale value of each pixel in a target image.
t(i,j) is a gray scale value of each pixel in a template in Sing
Library.
The most similarity of a target image and template in Sign Library
provides the minimum the Euclidean Distance (E (i,j)) compared
with the Euclidean Distance(E (i,j)) of every template in Sign
Library.
Z Zl
NED = —— J -
(imagesize )
(4)
To obtain the high accurate Road Sign position, the new
algorithm has been developed. The Data synchronization and the
basic principle of the Surveying are applied with RTK-GPS as
well as the constant from the sensor calibration. While the digital
CCD camera records the digital image of the Road Sign, the
moving rover station position are recorded by RTK-GPS
positioning techniques simultaneously. With the advantage of
RTK-GPS, the high accurate rover position is provided between
8-20 cm Standard Deviation every second. Moreover the
important factor is unavoidable that the rover station antenna is
on the road central line while the rover station is moving {see
Figure 5).
Fiaure 4 RTK-GPS method