TU Darmstadt / ULB / TUbiblio

Hypothesis Generation with Part Detectors for Multi-Target Tracking

Vancura, Daniel (2012)
Hypothesis Generation with Part Detectors for Multi-Target Tracking.
Technische Universität Darmstadt
Bachelorarbeit, Bibliographie

Kurzbeschreibung (Abstract)

Tracking people in occluded scenes is a hard problem and different approaches exist to offer more robustness against the effect of insufficient detections and clutter in retrieved positions. This thesis is about avoiding these drawbacks of missing hypotheses by using multiple detection categories for human tracking. Used for tracking is a system, that uses energy minimization techniques. The common approach for tracking is to use bounding boxes of full person detections only. The remaining problem with this approach is that partially occluded persons often cannot be detected when using a simple Histogram of Oriented Gradients (HOG) detector with a Support Vector Machine (SVM) for classification. To detect partially occluded pedestrians nonetheless, additional detectors can be run to detect only parts of persons that may appear in occluded scenes and enhance the tracking results. In this thesis the main steps and results of this approach will be evaluated and compared to existing approaches.

Typ des Eintrags: Bachelorarbeit
Erschienen: 2012
Autor(en): Vancura, Daniel
Art des Eintrags: Bibliographie
Titel: Hypothesis Generation with Part Detectors for Multi-Target Tracking
Sprache: Englisch
Publikationsjahr: 2012
Kurzbeschreibung (Abstract):

Tracking people in occluded scenes is a hard problem and different approaches exist to offer more robustness against the effect of insufficient detections and clutter in retrieved positions. This thesis is about avoiding these drawbacks of missing hypotheses by using multiple detection categories for human tracking. Used for tracking is a system, that uses energy minimization techniques. The common approach for tracking is to use bounding boxes of full person detections only. The remaining problem with this approach is that partially occluded persons often cannot be detected when using a simple Histogram of Oriented Gradients (HOG) detector with a Support Vector Machine (SVM) for classification. To detect partially occluded pedestrians nonetheless, additional detectors can be run to detect only parts of persons that may appear in occluded scenes and enhance the tracking results. In this thesis the main steps and results of this approach will be evaluated and compared to existing approaches.

Freie Schlagworte: Computer vision, Computer vision based tracking, Occlusion models, People detection, People tracking, Tracking
Zusätzliche Informationen:

42 p.

Fachbereich(e)/-gebiet(e): 20 Fachbereich Informatik
20 Fachbereich Informatik > Graphisch-Interaktive Systeme
Hinterlegungsdatum: 12 Nov 2018 11:16
Letzte Änderung: 12 Nov 2018 11:16
PPN:
Export:
Suche nach Titel in: TUfind oder in Google
Frage zum Eintrag Frage zum Eintrag

Optionen (nur für Redakteure)
Redaktionelle Details anzeigen Redaktionelle Details anzeigen