Geoscience Reference
In-Depth Information
traveling along a straight route will in general have a pattern that the directions of
the adjacent links connecting two GPS log points are similar. While turning across a
road section may result into signi
cant difference between the directions. Therefore,
we consider three variables as the sequence information, the connectivity
(Connectivity), the angle difference between the directions of two adjacent links
(AngleDiff), and the direction difference between the link connecting two nodes and
the matched road segment (DirectionDiff).
Here, the connectivity indicates whether two adjacent road segments are con-
nected each other. For example, to impute the probability of road segment i, we take
the previously matched road segment i
1 as a reference, and check whether the
current road i connects with the road i
1. This means we keep the information of
the previous matched road segment temporally to identify the potential road seg-
ment for the current location. In case that two adjacent locations identi
ed a same
road segment i, the connectivity is set as true.
The DirectionDiff can be calculated as mentioned previously. The angle dif-
ference (AngleDiff) is the absolute value of the difference between two adjacent
directions of the virtual links, as mentioned above. In case of the starting node, the
direction is set to 0. Therefore, the angle difference between the second link and the
previous link equals to the value of the direction with respect to the second GPS log
point.
In the model, we include two variables for the accuracy of the measurements, the
Position Dilution of Precision (PDOP) and the number of satellites (NSATS). The
variable PDOP is a measure of overall uncertainty of a GPS position, represents
the quality of GPS signals. A PDOP value of 1 indicates a good satellite con
g-
uration and high-quality data; conversely, PDOP values above 8 are considered
poor. The quality of the data decreases as the PDOP value increases.
The variable (DistToRoad) is the perpendicular distance from a node to a road
segment as presented above (Fig. 1 ). The variable of RoadAzimuth measures the
angle of the matched line object in a spherical coordinate system. Assume that a
vector from an origin to a point of interest is projected perpendicularly onto a
reference plane. Then the angle between the projected vector and a reference vector
on the reference plane is called the azimuth. Therefore, the azimuth information
indicates position of a road, which may provide useful input in combination with
other location variables.
The output variable is whether a road segment is the matched road. In this case,
it has two levels, yes or no. For each candidate road segment, we infer the prob-
ability through the conditional probabilities among input variables and the output
variable. For a set of found options, the road segment, which has the highest
probability, is taken as the rightly matched road.
Search WWH ::




Custom Search