The tracking component is an important part of an automotive environment perception system. The applied theoretical concepts as Kalman-Filtering are considered state-of-the-art and widely applicable in a standard manner. Recently, elementary new approaches emerged from military applications and particularly from RADAR based aerial surveillance, whose adaption to automotive applications seems worthwhile. This contribution shortly introduces these new methods, their theoretical background and derived algorithms for simultaneous detection and tracking of an unknown time-changing number of objects. Further, an automotive application for Adaptive-Cruise-Control functions (ACC) using one of these derived algorithms is presented. JIPDA allows the incorporation of direct measurements of target existence in a probabilistic and therefore sensor and signal processing independent way instead of only relying on monotonic appearance of measurements in a certain region for track confirmation. State constraints and prior knowledge about target existence can be integrated seamlessly via the process model for target existence as shown for digital map information, occlusion reasoning and field of view constraints. The detection decision can be forced by a single threshold operation on the posterior existence probability. In the data association domain, JIPDA outperforms Greedy approaches like the nearest neighbour method or global approaches like the Kuhn-Munkres algorithm. Traditional Greedy or global methods based on Mahalanobis-distances (MHD) tend to false association decisions because these tracks take away measurements from the true tracks. In JIPDA the spatial arrangement encoded in MHDs, the existence probabilities of the tracks, and the existence probabilities of the measurements all influence the association weights and therefore the mentioned error types are avoided. The system is implemented in a demonstration vehicle with serial night vision camera and a multi beam Lidar sensor. Even standing vehicles are detected in distances up to 150 meters and the system can keep track of already detected vehicles up to 200 meters. The current development phase allows non-synthetic in-vehicle demonstrations in real traffic scenarios.


    Zugriff

    Zugriff über TIB

    Verfügbarkeit in meiner Bibliothek prüfen


    Exportieren, teilen und zitieren



    Titel :

    Multi-object-tracking for advanced driver assistance systems using random finite set theory


    Beteiligte:


    Erscheinungsdatum :

    2008


    Format / Umfang :

    14 Seiten, 9 Bilder, 1 Tabelle, 14 Quellen


    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Datenträger


    Sprache :

    Englisch




    Multi-object-tracking for advanced driver assistance systems using random finite set theory

    Maehlich,M. / Dietmayer,K. / Daimler,Stuttgart,DE et al. | Kraftfahrwesen | 2008



    Advanced driver-assistance system using accurate object proposals by tracking detections

    SCHULTER SAMUEL / CHOI WONGUN / SINGH BHARAT et al. | Europäisches Patentamt | 2019

    Freier Zugriff

    ADVANCED DRIVER-ASSISTANCE SYSTEM USING ACCURATE OBJECT PROPOSALS BY TRACKING DETECTIONS

    SCHULTER SAMUEL / CHOI WONGUN / SINGH BHARAT et al. | Europäisches Patentamt | 2018

    Freier Zugriff

    Road tracking using particle filters for Advanced Driver Assistance Systems

    Chandran, Prashanth / John, Mala / Santhosh Kumar S et al. | IEEE | 2014