Videobasierte modellgestützte Objekterkennung (Volume 28) (English shop)

Preview

Table of Contents, Datei (54 KB)
Extract, Datei (130 KB)

To meet the increasing demands of future driver assistance systems on the quality of information about the local environment of the ego vehicle, this thesis develops algorithms for video-based vehicle and lane detection. The methods for vehicle detection combine expert knowledge with modern machine learning techniques in an overall system that can be adapted to various problems of video-based object recognition. The use of specific sensor models enables the subsequent multi-sensor tracking of vehicles in a Kalman filter based object tracker.

An algorithm for video-based lane tracking is extended by measures that increase robustness, thus enabling automatic lane guidance of the test vehicle even in complex driving scenarios.