Methodology
It is important to know where unpaved roads are and what
proportion of the road network is unpaved, both for transpor-
tation asset management and
UAV
mission planning require-
ments. Not all counties in the southeastern Michigan study
area (Figure 2) have an accurate inventory of their unpaved
road location and an unpaved versus paved attribute is not
found in the State of Michigan Geographic Framework roads
data. In many Michigan counties, unpaved roads comprise
a significant proportion of the overall road network which
makes an accurate inventory important. Oakland County in
southeastern Michigan, which has mixed urban/rural land
use, estimated it has approximately 1,200 kilometers of un-
paved roads out of a total of more than 4,800 kilometers that
the county Road Commission for Oakland County is respon-
sible for maintaining
Gravel_Roads.aspx
). This distance is greater than the total
road network length in some counties in the Upper Peninsula
of Michigan. By contrast, Wayne County (part of
SEMCOG
),
which contains the city of Detroit, has few unpaved roads, and
its county road network was not processed for this project.
Identifying Unpaved Roads in a County Road Network
Identifying and attributing unpaved road segments in the
Michigan Framework roads
GIS
layer is a multi-step process.
Only roads that are National Functional Classification (
NFC
)
types 4 (Minor Arterials), 5 (Major Collectors), 6 (Minor Col-
lectors), or 7 (Local) were assessed as they can be either paved
or unpaved. Interstates (
NFC
type 1), Other Freeways (2), and
Other Principal Arterials (3) are nearly always paved and are
excluded from further analysis.
To identify unpaved road segments within a county road
network, the shapefile containing unpaved road polygons ex-
ported from eCognition
®
is loaded into ArcGIS
®
. The imported
polygons are then intersected with the Michigan Framework
roads shapefile to create a line feature that represents the
road segments that overlay a polygon classified as unpaved in
eCognition.
The attribute table for the segmented unpaved roads shape-
file exported from eCognition is where the changes will be
made to support adding the paved/unpaved attribute for each
road segment. The length of each road segment is determined
Figure 1. An RGB example of the RGBIR aerial images from Oakland County MI that were processed using eCognition and
ArcGIS to map the location of unpaved versus paved roads in SE Michigan. A = unpaved road dominated by natural aggre-
gate; B = unpaved road constructed with crushed limestone; C = paved asphalt road.
Figure 2. Study area in southeastern Michigan for unpaved
roads mapping for inventory and mission planning inputs.
378
May 2017
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
