Search

International audience; This article deals with the detection of pedestrians by means of a laser sensor. This sensor placed on the front of a vehicle collects information about distance distributed according to 4 horizontal planes. In order to improve the robustness of pedestrian detection using a single laser sensor we propose here a detection system based on decentralized fusion of information located in the 4 horizontal laser planes. A Parzen kernel method is described and allows to extract "pedestrian objects" in each laser layer before to carry out a decentralized fusion based also on the Parzen kernel method. Many experimental results validate and show the relevance of our pedestrian detection algorithm in regard to a method using only a single-row laser-range scanner.

International audience; This paper introduces an implementation of the Polynomial Extended Kalman Filter (PEKF) to solve the Simultaneous Localization and Map building (SLAM) problem. The proposed solution is a filtering algorithm which is a polynomial transformation of state evolution and measurement equations. The performances of the algorithm have been evaluated through two simulation runs. The first ones underline consistency improvement in comparison with the standard Extended Kalman Filter. The other simulation results show the PEKF efficiency when the values of measurement noises are high. At the end, experiments with Victoria Park data are presented too.

International audience; Cet article concerne la détection des piétons à l'aide d'un capteur laser. Ce capteur placé à l'avant d'un véhicule recueille des informations distance réparties selon quatre nappes. Au même titre qu'un véhicule, un piéton constitue dans un environnement de conduite un obstacle qu'il faut détecter, localiser puis identifier et éventuellement suivre. Une synthèse bibliographique concernant les travaux relatifs à la détection des piétons à l'aide de données LIDAR est présentée. Une méthode de détection, d'identification puis de suivi de piétons à partir de cet unique capteur laser quatre plans est proposée et discutée. L'algorithme de détection regroupe les points en différentes classes géométriques dans le but de filtrer le fond de l'image (murs, voitures, panneaux, etc.). Afin d'isoler les piétons, nous proposons une méthode non-paramétrique exploitant la technique de fenêtrage de Parzen. Finalement, pour caractériser au mieux la trajectoire complexe d'un piéton, le pistage repose sur l'utilisation classique d'un filtre à particules. Les résultats obtenus sur des données réelles pour plusieurs scénarios sont présentés.

International audience

International audience; This paper deals with evaluation of centralized fusion for two dissimilar sensors for the purpose of road obstacle tracking. The aim of sensor fusion is to produce an improved state estimate of a system from a set of independent data sources. Indeed, for a robust environment perception, see as obstacles here, several sensors should be installed in the equipped vehicle: camera, lidar, radar, etc. In our case, the motivation for this work comes from the need to track road targets with lidar measurements combined to radar ones. Thus, the aim is to combine effectively radar range measurements (i.e. range and range rate) with Lidar Cartesian measurements for a ”turn” scenario. Centralized fusion, i.e. measurement fusion, for two dissimilar sensors is considered here for evaluation. Evaluation is based on Cramer-Rao Lower Bound (CRLB) which is the basic tool for investigating estimation performance as it represents a limit of cognizability of the state. In the target tracking area, a recursive formulation of the Posterior Cramer- Rao Lower Bound (PCRLB) is used to analyze performance. Many bound comparisons are made according to used scenarios and various sensors configurations. Moreover, two algorithms for target motion analysis are developed and compared to the theoretical bounds of performance: the extended Kalman filter and the particle filter.