A strategy for matching stereo pairs using feature similarity has been developed using the SUSAN corner detector to locate the feature primitives. The use of stereo vision provides information regarding a vehicle's position on the road and allows each object to be clearly distinguished. Using a hypothesis and test strategy, instantiation of models into a scene allows automatic identification of visible and partly visible features. Using a user defined ranking system an image can be tested for distinguishing features. The region of the scene in which to test for the features is defined by the model instantiation. Perspective distortion and camera to viewscreen transformation allows the features to be instantiated in a realistic fashion. The instantiated features can therefore be regarded as templates for matching purposes.
Vehicle classification using stereo vision
1997-01-01
5 pages
Aufsatz (Zeitschrift)
Englisch
image classification , camera to viewscreen transformation , road vehicles , stereo image processing , road vehicle classification , automatic identification , identification , template , matching stereo pairs , hypothesis and test strategy , feature similarity , SUSAN corner detector , user defined ranking system , feature primitives , stereo vision , partly visible features , visible features , feature extraction , position , image matching , distinguishing features , perspective distortion , instantiation
Stereo vision based vehicle detection
Tema Archiv | 2010
|Stereo Vision-based Vehicle Detection
British Library Conference Proceedings | 2000
|Stereo vision-based vehicle detection
IEEE | 2000
|ARPA Unmanned Ground Vehicle Stereo Vision Program
NTIS | 1994
|