Optische Kfz-Innenraumüberwachung - Optical vehicle compartment surveillance
The optical monitoring of the automobile passenger interior compartment is concerned with applications based on optical coverage and image processing in order to provide various functionalities. The emphases of this thesis are optical occupant state recognition and optical interior protection. The interior protection is an integral part of the car alarm system and must be able to detect unauthorized entry. A significant judgement criterion apart from the high detection rate of penetrating objects is a low false alarm rate, which should be achieved using a high spatial resolution of an optical system. Additionally the power dissipation of car alarm systems must be extremely low. This can be obtained when employing modern CMOS image sensors. The task of occupant state recognition arose from the unresolved problem of rear facing infant seat detection and Out-of-Position detection. Tragic accidents involving rear facing infant seats placed in the passenger seat have exemplified the necessity for the detection of infant seats and other critical situations. Optical occupant state recognition encompasses seat occupancy, infant seat detection, and Out-of-Position detection, which up until now have been realized separately - if at all. The critical requirements of this application are adaptability to strongly varying environmental conditions and fast response time. Optical systems for use in passenger compartment monitoring have not yet been the subject of ex tensive research. By employing cameras based on CMOS image sensors the optical stimuli within the vehicle are measured. In order to enable several applications with one system, the position of the camera(s) and the optical influences at this position are examined. In the case of the optical interior protection, algorithms have to be developed enabling the detection of the penetration of objects. Due to the restriction of low power dissipation, these algorithms should not be too complex. The integration of on-chip preprocessing and other electronic features can be combined to make CMOS image sensors power efficient. For optical occupant state recognition concepts with a low cost of computation have to be developed, which are also capable of working in conjunction with the interior monitoring. Thus a high frame rate has to be obtained with these systems. In addition algorithms are necessary that can determine the status of each respective seat. Since the optical occupancy state recognition is a security-sensitive application, additional algorithms are examined that detect very accurately (up to object recognition) the occupation state of the vehicle.
Advisor:Prof. Bedrich J. Hosticka, Ph.D.
School:Universität Duisburg-Essen, Standort Essen
Source Type:Master's Thesis
Keywords:elektrotechnik gerhard mercator universitaet
Date of Publication:05/31/2001