TY  - JOUR
PY  - 2007//
TI  - Combination of Feature Extraction Methods for SVM Pedestrian Detection
JO  - IEEE transactions on intelligent transportation systems
A1  - Parra, I
A1  - Llorca, DF
A1  - Sotelo, MA
A1  - Bergasa, LM
A1  - Garrido, Miguel Aangel Garcia
A1  - Ocana, M
A1  - Revenga de Toro, P
A1  - Nuevo, J
SP  - 292
EP  - 307
VL  - 8
IS  - 2
N2  - In this paper the authors provide an automated feature extraction mechanism using stereovision and a support vector machine (SVM) for pedestrian detection in Intelligent Transportation Systems (ITS). A monocular detection method is used, as such systems are more easily produced en mass for market purposes. Pedestrian detection is also difficult for such applications because it is conducted from a moving vehicle, must quantify and select many candidates in large groups called multi-candidate (MC) generation, and must be able to maintain such functionality with or without full daylight. In order to reduce the number of false candidates in a given group the model also employs subtractive clustering. The authors propose that, for future projects, a more robust motion-based and position-dependent incorporative mechanism is used to enhance the detection method presented here.<p />
LA  - 
SN  - 1524-9050
UR  - http://dx.doi.org/
ID  - ref1
ER  -