Skip to main content
. 2012 Aug 2;12(8):10407–10429. doi: 10.3390/s120810407

Table 1.

Performance results for the route detector. N is the number of points per trajectory.

Video Description N Duration (s) Processing Time (s) Number of Objects Number of Paths Number of Paths Identified Number of wrong paths identified % identified paths % wrong paths
A synthetic video of a highway with two lanes, one direction per lane. 5 600 318s 359 2 2 2 100% 100%
10 600 283s 359 2 2 0 100% 0%
20 600 402s 359 2 2 0 100% 0%
A synthetic video of a highway with four lanes, two lanes for each direction. 5 480 26s 58 4 3 1 75% 25%
10 480 52s 58 4 4 0 100% 0%
20 480 123s 58 4 4 0 100% 0%
A real video of a complex intersection, involving three roads for cars with two directions, and several sidewalks. 5 180 118 191 6 (cars)
7 (people)
6 (cars)
2 (people)
1 (cars)
1 (people)
100% (cars)
28% (people)
16% (cars)
14% (people)
10 180 284 191 6 (cars)
7 (people)
6 (cars)
2 (people)
1 (cars)
1 (people)
100% (cars)
28% (people)
16% (cars)
14% (people)
20 180 1696 191 6 (cars)
7 (people)
5 (cars)
2 (people)
2 (cars)
1 (people)
86% (cars)
28% (people)
33% (cars)
14% (people)
Part of the MIT video benchmark showing a complex intersection with roads and sidewalks. 5 202 196 145 6 (cars)
6 (people)
3 (cars)
1 (people)
3 (cars)
1 (people)
50% (cars)
16% (people)
50% (cars)
16% (people)
10 202 831 145 6 (cars)
6 (people)
4 (cars)
1 (people)
2 (cars)
1 (people)
66% (cars)
16% (people)
33% (cars)
16% (people)