Mile-marker
background
Figure 7: Tracing the edges of the mile-marker from the
bottom to the top.
5.3 Extraction of the Log-Mile Number
While the edges are traced from the bottom of the mile-
marker to the top, the average and the standard deviation of
the pixels between the detected edge points are computed.
As we reach the number at the top of the monument, the
gray-value average decreases and the standard deviation
gets larger for a number of lines (figure 8). These lines of
lower gray values, but higher standard deviation define the
area in which the log-mile number is written; they are
chosen to identify the position of this number. It can be
read by the OCR techniques described below.
ae Top of mile-marker
Lines of the image of lower
average gray-values and
larger standard deviations
A
Figure 8: Locating the log-mile number at the top of a
mile-marker by computing average and standard
deviation of the gray-values between the edges.
It is based on a simple algorithm which allows to
identify characters from vector graphics. For this technique
the numbers must appear upright in the images, which is
the cased in our mile-marker photos. A prerequisite is the
vectorization of the number, e.g. by thresholding the image
and thinning the black (binarized) digits to a width of only
one pixel. By chain-coding a polyline is created which
forms the detected number. It is smoothed, in order to
eliminate noise from binarization.
Each polyline (= number) is intersected by some
horizontal and vertical lines; the number of intersections in
x and y directions is counted (figure 9). Each digit is
uniquely identified by the number of intersections, which
are stored for comparison purposes in the program. The
major advantage of this technique is its computational
speed.
124
AN
intersection points
F igure 9: A character can be identified in a unique way by
counting intersections of horizontal and vertical lines,
e.g.number of intersections per line: line I : 3
line 2 : 3
line 3 : 2
line 4 : 1
line 5 : 1
6. CONCLUSIONS
In this article we tried to show the big potential of using
digital image-pairs for automatic mapping of features from
land-based vehicles. Although we are currently limiting
ourselves to a few well-defined objects, this system could
be applied to create a complete highway or railroad
inventory, containing traffic signs, mile markers, and road
signals. Due to the known position and attitude of the
vehicle, any object identified in an image-pair is
immediately available in a global coordinate system.
As both the locations of image-pairs, and the detected
objects are given in geographic coordinates, it makes sense
to integrate these functions in a GIS. We are currently
working on this problem, combining an existing vector GIS
with the stereo-image data-base. The feature extraction
functions serve as tools for collecting attribute information
from the imagery. In order to allow the operator to better
check the results a stereo-monitor will be used to display
image-pairs and to select objects directly in 3-dimensions.
Once a set of image analysis functions is available we
envision to install image processing capabilities directly in
the vehicle. Together with real-time, differential GPS we
will be able to extract road features while driving along
highways. Then we can truely talk about real-time
mapping.
7. REFERENCES
Bossler J., Goad C., Johnson P., Novak K., 1991. *GPS and
GIS Map the Nation's Highways." GeoInfo
Systems Magazine, March issue, pp. 26-37.
Goad C., 1991. “The Ohio State University Highway
Mapping System: The Positioning Component.”
Proceedings of the Institute of Navigation
Conference, Williamsburg, VA, pp. 117-120.
He G., Novak, K., 1992." Automatic Analysis of Highway
Features from Digital Stereo-Images."
International Archives of Photogrammetry and
Remote Sensing, Vol. , Commission III.
Novak K., 1991. “The Ohio State University Highway
Mapping System: The Stereo Vision System
Component.” Proceedings of the Institute of
Navigation Conference, Williamsburg, VA, pp.
121-124.
t + © ee 3 er FH OO AB & À ©
In
A--— "^^^ bid. | Ey 5 Lid L2 fV ffi
od Los
2 d eed Aaa FA^
