ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS’’, Bangkok, May 23-25, 2001
245
4 Result
The algorithm has been valiadted in this study below:
Some example of Sign Recogntion Result form the study
Figure 8 Sign No.4 Result
-Four Points were selected to calculate the affine parameter and
the distorted road sign image was extracted.
-After a distorted sign image was selected and corrected for
geometry, two images were displayed. The left image is a
geometric corrected image of the same size as the sign in Sign
Library. The right image is an interpolated image.
-After a geometric corrected image was calculated for Normalized
Euclidean Distance with every Sign image in Sign Library, the
program sorted the minimum Normalized Euclidean Distance
(NED) from all images of Sign Library and finally that image was
display as the temple matching result. (See Figure 8)
Figure 9 the Sign No.4 Recognition Result
The Sign No.4 Recognition Result
S c
I1
z m
0.002
N W y
= « c 0.0018
,r «
£ 0.0016
□
0.0014 rrTT n-rrp—
O o'
TTrrrn-rrrT-f"TTTr-J T r; r
oo
The Number of Sign in Sign Library
o The Normalized Euclidean Distance j
