(a)
(b)
(c)
Surveillance cameras capture the videos of moving vehicles and the vehicle blobs are extracted by background subtraction. The extracted vehicle blobs convey the information of the vehicle shapes as shown in Figure 1.
|
(a) |
|
(b) |
|
(c) |
Figure 1. (a) A sedan and its extracted blob; (b) A mini-van and its extracted blob; (c) A SUV and its extracted blob.
Train a neural networks with the extracted vehicle blobs and the corresponding vehicle classes. The trained neural networks can fulfill the task of classify vehicles as shown in Figure 2.
(a) |
(b) |
(c) |
Figure 2. The classification results of the trained neural networks on a sedan, a mini-van, and a SUV. (a) The type of sedan got a score of 0.99492, much higher than those of others; (b) The type of mini-van got a score of 0.22133, higher than those of others; (c) The type of SUV got a score of 0.89979, much higher than those of others. (*note: the highest score is 1 and the lowest score is 0)