Geoscience Reference
In-Depth Information
Table 1 Experiments results
Test #
Duration
(min)
km
Average
speed
IAPT
(%)
NAI
NP
NFP
POI detected
1.1
3.45
1
25
100
122
6
0
Yes
1.2
3.40
1
35
100
124
7
1
Yes
1.3
3.40
1
40
100
129
6
1
Yes
2
3.25
1
35
100
268
2
0
Yes
3.1
2.15
1.2
30
90
72
5
0
Yes
3.2
2.25
1.2
30
90
76
6
1
Yes
3.3
2.05
1.2
35
90
69
4
0
Yes
4
9.55
2.5
35
18
102
6
0
Yes
No
a
5.1
2.05
1.2
35
90
70
1
1
5.2
2.15
1.2
30
90
70
0
0
No
6
9.40
2.5
30
34
187
5
0
Yes
7
3.50
1.2
25
100
107
10
0
Yes
8
3.25
1.2
25
90
76
5
0
Yes
a
This false positive was not expected
that route (was a regular false positive). This false positive detection was not seen in
the other repetition performed in the same route. It is important to refer that the false
positives in 1.2, 1.3, 3.2 and 5.1 were detected at a considerable distance of the
intended POI (from 180 to 220 m).
6 Discussion
The analysis of the achieved results makes it possible to verify that, for all the
performed tests, the target POI was successfully detected in time, and more than
once per repetition.
In the
first experiment (1.1, 1.2 and 1.3), the POI was detected 6, 7 and 6 times
respectively, but also two false positives were found. The false positives were
detected with the vehicle stopped at a semaphore, which discards the possibility of
being caused by the road conditions or velocity. In the second experiment (2), a
single image was available in the database (the POI only had one image associate)
for the image analysis matching. For this reason there were performed a lot of more
matches than in other experiments. On the other hand, only two positives were
detected, because the road where the POI is visible for the camera is only around
50 m long (short time available for detection). In the third experiment (3.1, 3.2 and
3.3), only 90 % of the route was covered by image analysis, because the 10 %
remaining were off the analysis maximum distance perimeter. This allowed us to
verify that the combination of both technologies was working well, and that the
image analysis was being triggered in the right time. Only one false positive was
Search WWH ::
Custom Search