Skip to Main content Skip to Navigation
Conference papers

Vers une détection de piétons temps réel par apprentissage de forme dans l'image de profondeur

Abstract : This paper presents a method for detecting pedestrians in a depth image, established from a pair of calibrated cameras (stereoscopic head). We propose to learn the characteristics of a pedestrian thanks to a boosting type algorithm, using weak classifiers created from simple statistical characteristics of distance within a sub-window of the area of analysis. These assumptions will then be confirmed by a second, classic detector based on analysis of the visual appearance of pedestrians. Our resulting method allows for the detection of pedestrians at a rate close to real time, using the concept of integral image, applied to the calculation of 3D descriptors. The proposed method is compared with a traditional method for detecting vertical obstacles on a real annotated video sequence. Our method reduces the number of false positives by approximately 60% compared to an obstacle-detection method, while reducing the computation time.
Document type :
Conference papers
Complete list of metadata

Cited literature [10 references]  Display  Hide  Download
Contributor : Peter Sturm Connect in order to contact the contributor
Submitted on : Friday, July 22, 2011 - 11:28:49 AM
Last modification on : Monday, December 13, 2021 - 9:16:01 AM
Long-term archiving on: : Monday, November 12, 2012 - 3:10:08 PM


Files produced by the author(s)


  • HAL Id : inria-00610513, version 1



Loïc Jourdheuil, Nicolas Allezard, Thierry Chateau. Vers une détection de piétons temps réel par apprentissage de forme dans l'image de profondeur. ORASIS - Congrès des jeunes chercheurs en vision par ordinateur, INRIA Grenoble Rhône-Alpes, Jun 2011, Praz-sur-Arly, France. ⟨inria-00610513⟩



Les métriques sont temporairement indisponibles