GRIDVECTOR (version 1) : AN ARC/INFO AML PROGRAM TO EXTRACT
LINEAR FEATURES FROM A GRAY-SCALE IMAGE OF A PAPER
GEOLOGICAL MAP

by

Feliks Persits

 

Open-File Report 97-713


Previous Next

The following figures show some examples.

Figures 1-5 show the GRIDVECTOR application on a fragment of a 1:7,500,000 scale geologic map of relatively poor quality (i.e. pale colors, variable line width, very tiny polygons). Figure 1 shows the input grid with arrows pointing to latitude/longitude lines that are considered to be noise. Other noise features are text labels and a background dot pattern. The final line coverage (after GRIDLINE) in figure 4 shows advantages of the GRIDVECTOR algorithm. Coordinate lines and many of the text labels are completely filtered out. Some of the polygons look completed and are ready for geologic attributing. The results in the lower-left corner are of poor quality; further editing and correcting need to be done before geologic attributing can start because there are many small polygons and small distances between lines in comparison to the GRID filter size. At the same time, in figure 3 (binary image after grid processing which serves as the input for GRIDLINE) the white lines are more contiguous and more like the original. This illustrates the fact that the effectiveness depends on GRIDLINE. Figure 5 shows another run on the same grid, using different parameters; - a higher threshold for dark lines and average slope. There are fewer false lines in this coverage, but at the same time more digitization will be needed. Figure 2 shows an intermediate average slope grid where the white pixels are maximums (arrows point to some). The coordinate lines and many of the text labels have disappeared but much correction and digitization is still needed. Half of the desired lines are present and therefore the same amount of handwork has been saved.

The second example shows another GRIDVECTOR application. The goal was to select and trace sea depth contour lines from a gray-scale image (Figure 6) in which areas between the contours are filled in by different patterns. In spite of this image noise, lines on the final coverage (Figure 7) follow the original contours accurately. Minor corrections, like filling in gaps and removing false lines, will still need to be done.

Conclusion:

GRIDVECTOR has been created as a helpful tool for paper map digitization, that continues to be a most tedious and time-consuming task in new GIS projects. The standard ARC/INFO GRIDLINE or ARCSCAN subsystem can be applied only to binary images and therefore requires handmade mylar copies of maps. GRIDVECTOR provides a more efficient gray-scale scanning method of moving a paper map into ARC/INFO format.

Figure 1. A fragment of a 1:7,500,000 scale geologic map
FIG. 1 – A fragment of a 1:7,500,000 scale geologic
map of relatively poor quality, i.e. pale colors,
variable line width, tiny polygons. Gray-scale
scanned image of this fragment has been converted
to grid by IMAGEGRID command and used as
input for GRIDVECTOR program. Arrows point to a
latitude/longitude lines that are considered to be
noise. Other noise features are text labels and
background dot patterns.
Figure 2. An intermediate average slope grid
FIG. 2 – An intermediate average slope grid, from
map fragment in figure 1, where the white pixels are
maximums. Latitude/longitude grid lines and many
of the text labels have disappeared.
Figure 3. Binary grid, from map fragment in figure
1
FIG. 3 – Binary grid, from map fragment in figure 1,
after completion of input grid processing which
serves as the input for GRIDLINE. Noise features
including latitude/longitude grid lines are
converted to groups of speckles that are in turn
removed by the GRIDDESPECKLE command of
ARC/INFO.
Figure 4. Final line and polygon ARC/INFO <BR>
coverage after GRIDLINE
FIG. 4 – Final line and polygon ARC/INFO
coverage after GRIDLINE, from map fragment in
figure 1. Coordinate lines and many of the text
labels are completely filtered out. Some of the
polygons look completed and are ready for geologic attributing. The results in the lower-left corner
however are of poor quality and further editing a
nd correcting needs to be done before geologic
attributing can start.
Figure 5. Another run on the same grid, as figure 1
FIG. 5 – Another run on the same grid, as figure 1,
using as different parameters: a higher threshold for
dark lines and average slope. There are fewer false
lines in this coverage, but at the same time more
digitization is needed. Map fragment from figure 1.
Figure 6. A fragment of a map at 1:7,500,000
FIG. 6 – A fragment of a map at 1:7,500,000
scale showing sea depth contours that are
to be extracted. Areas between the contours
are filled in by different patterns.
Figure 7. Final line ARC/INFO coverage
FIG. 7 – Final line ARC/INFO coverage
of the map fragment, figure 6. Minor
corrections, like filling in gap and removing
false lines, will still need to be done.

Previous Next

U.S. Geological Survey Open-File Report 97-713


USA.gov logo  Take Pride in America button