Geoscience Reference
In-Depth Information
4 Testing and Results
In order to properly evaluate the CV-GPS method, a set of experiments was
performed, evaluating if the proposed method can provide geolocation improve-
ments, and if the image analysis system is working as expected in different light
conditions (which varies, for instance, with the atmospheric conditions), vehicle
velocities or road pavement conditions. It is expected, though, that the system fails
in situations where is not possible to get a clear picture of the elements to detect
(for instance, under heavy rain). The next sections present the test scenarios used
for evaluating the system and the routes and Points of Interest used for performing
the experiments. After that, the obtained results are presented and discussed.
4.1 Test Routes and Points of Interest
In the development of the test scenarios three test routes were used, all in the Porto
downtown, Portugal. The routes were chosen according with speci
c characteristics
like different distances, road pavements, illumination conditions, and the distance
and line of visibility from the road to the Points of Interest expected to be detected.
Also three POIs were used, speci
cally, POI1 (GPS coordinates 41.145669,
8.614728, two model images associated in the database), the POI2 (GPS coor-
dinates 41.149594,
-
8.610303, two model images associated in the database) and
POI3 (GPS coordinates 41.1475314,
-
8.6165759, one model image associated in
the database). These POIs have distinct shape and textures and as we can see
in Fig. 5 , their model images were captured under less favourable light conditions
(in order to test the system under the worst conditions).
-
4.2 Experiments
To correctly validate the system, eight experiments were performed, some of them
inclusively repeated, in a total of 13 tests. In order to provide enough information to
Search WWH ::




Custom Search