Visual Scene Understanding
My current research focus is on generic object recognition methods and models for vision-based 3D scene understanding in complex urban environments.
The main research during my PhD involved statistical models of pedestrian shape and appearance with an application to on-board pedestrian detection for intelligent vehicles. Rather than deriving an explicit model of human appearance, we follow a pattern classification approach, where appropriate models are implicitly learned from a large number of training examples.
This research and corresponding implementation is the main foundation of the Daimler vision-based pedestrian detection system which is available in the Mercedes-Benz E-Class and S-Class models in the context of Mercedes-Benz Intelligent Drive.
Keywords: Pedestrian recognition, scene understanding, image processing, machine learning, intelligent vehicles