An Adaptable Object Classification Framework
Stefan Wender, Klaus Dietmayer · 2006
The classification of observed objects in the vehicle's environment is necessary for several active safety systems. A framework for the object classification task is introduced. The classification benefits from pattern classification as well as from rule based a priori knowledge. The framework can serve different applications at the same time. A new approach is applied to adapt the output for each application to its special requirements. This adaptation consumes only very little processing time and can be performed for multiple applications without affecting the framework's real time properties. The practical usage of the framework is illustrated by the classification of measurements of a laser scanner, but the framework is also applicable for other types of sensors