The benefits of dense stereo for pedestrian detection

• Verfasst von: Keller, Christoph Gustav [VerfasserIn]
• Verfasst von: Enzweiler, Markus [VerfasserIn]
• Verfasst von: Rohrbach, Marcus [VerfasserIn]
• Verfasst von: Llorca, David Fernández [VerfasserIn]
• Verfasst von: Schnörr, Christoph [VerfasserIn]
• Verfasst von: Gavrila, Dariu M. [VerfasserIn]
• Titel: The benefits of dense stereo for pedestrian detection
• Verf.angabe: Christoph G. Keller, Markus Enzweiler, Marcus Rohrbach, David Fernández Llorca, Christoph Schnorr, Dariu M. Gavrila
• E-Jahr: 2011
• Jahr: 12 May 2011
• Umfang: 11 S.
• Fussnoten: Gesehen am 28.07.2022
• Titel Quelle: Enthalten in: Institute of Electrical and Electronics EngineersIEEE transactions on intelligent transportation systems
• Ort Quelle: New York, NY : Inst. of Electrical and Electronics Engineers, 2000
• Jahr Quelle: 2011
• Band/Heft Quelle: 12(2011), 4, Seite 1096-1106
• DOI: doi:10.1109/TITS.2011.2143410
• Datenträger: Online-Ressource
• Sprache: eng
• Sach-SW: Active safety
• Sach-SW: Cameras
• Sach-SW: computer vision
• Sach-SW: Feature extraction
• Sach-SW: intelligent vehicles
• Sach-SW: Intelligent vehicles
• Sach-SW: pedestrian detection
• Sach-SW: Stereo vision
• Sach-SW: Support vector machines
• K10plus-PPN: 1811893724

Abstract

This paper presents a novel pedestrian detection system for intelligent vehicles. We propose the use of dense stereo for both the generation of regions of interest and pedestrian classification. Dense stereo allows the dynamic estimation of camera parameters and the road profile, which, in turn, provides strong scene constraints on possible pedestrian locations. For classification, we extract spatial features (gradient orientation histograms) directly from dense depth and intensity images. Both modalities are represented in terms of individual feature spaces, in which discriminative classifiers (linear support vector machines) are learned. We refrain from the construction of a joint feature space but instead employ a fusion of depth and intensity on the classifier level. Our experiments involve challenging image data captured in complex urban environments (i.e., undulating roads and speed bumps). Our results show a performance improvement by up to a factor of 7.5 at the classification level and up to a factor of 5 at the tracking level (reduction in false alarms at constant detection rates) over a system with static scene constraints and intensity-only classification.