Samir Hachour - Academia.edu (original) (raw)

Papers by Samir Hachour

Multi-target tracking with credal classification and kinematic data

This article proposes a method to classify multiple maneuvering targets at the same time. This ta... more This article proposes a method to classify multiple maneuvering targets at the same time. This task is a much harder problem than classifying a single target, as sensors do not know how to assign captured measurements to known targets. This article extends previous results scattered in the literature and unifies them in a single global framework with belief functions. Through two examples, it is shown that the full algorithm using belief functions improves results obtained with standard Bayesian classifiers and that it can be applied to a large variety of applications.

In this paper, we study the problem of joint tracking and classification of several targets at th... more In this paper, we study the problem of joint tracking and classification of several targets at the same time. Targets are considered to be known and sufficiently separated so that they cannot be confused. Our goal is to propose a full methodology that is robust to missing information. The classical probabilistic approach with Bayesian tools is improved with belief functions. A simulation concerning the identification of go fast boats in a piracy problem shows that our approach improves previous results.

A Belief Function Solution for Stator Insulation Robustness Study

2019 9th International Conference on Power and Energy Systems (ICPES), 2019

This paper proposes a model-based decision taking solution for electrical machines winding insula... more This paper proposes a model-based decision taking solution for electrical machines winding insulation robustness study. The solution is based on the Belief Function (BF) theory. It is processed in two main steps: a first one aims to learn Weibull model parameters from some labeled aging Partial Discharge Inception Voltage (PDIV) data. Then a second classification step separates some unlabeled PDIV data according to the learnt Weibull models. The classification results can give information on the robustness and the reliability of the Electrical Insulation System (EIS) under a thermal constraint.

In this paper, a proposition is made to learn the parameters of evidential contextual correction ... more In this paper, a proposition is made to learn the parameters of evidential contextual correction mechanisms from a learning set composed of soft labelled data, that is data where the true class of each object is only partially known. The method consists in optimizing a measure of discrepancy between the values of the corrected contour function and the ground truth also represented by a contour function. The advantages of this method are illustrated by tests on synthetic and real data.

This paper presents a distributed solution for multi-object tracking and classification. The stat... more This paper presents a distributed solution for multi-object tracking and classification. The state of objects is partially observed by a set of sensors organized in a network. The idea is to exchange partial data throughout the network and provide a complete information at each sensor level. The proposed solution involves a finite time average consensus where existing solutions are based on asymptotic consensus. The consensus algorithm intervenes in both distributed tracking and classification of multiple objects. It is firstly used to complete information about objects trajectories and secondly to complete beliefs concerning the classification. Simulation results show the relevance of the proposed solution.

Cette these aborde le probleeme du suivi et de la classification de plusieurs objets simultanemen... more Cette these aborde le probleeme du suivi et de la classification de plusieurs objets simultanement.Il est montre dans la theese que les fonctions de croyance permettent d'ameliorer les resultatsfournis par des methodes classiques a base d'approches Bayesiennes. En particulier, une precedenteapproche developpee dans le cas d'un seul objet est etendue au cas de plusieurs objets. Il est montreque dans toutes les approches multi-objets, la phase d'association entre observations et objetsconnus est fondamentale. Cette these propose egalement de nouvelles methodes d'associationcredales qui apparaissent plus robustes que celles trouvees dans la litterature. Enfin, est abordee laquestion de la classification multi-capteurs qui necessite une seconde phase d'association. Dans cedernier cas, deux architectures de fusion des donnees capteurs sont proposees, une dite centraliseeet une autre dite distribuee. De nombreuses comparaisons illustrent l'interet de ces travau...

Belief Function Based Multisensor Multitarget Classification Solution

Multisensor Data Fusion

Improving an Evidential Source of Information Using Contextual Corrections Depending on Partial Decisions

Belief Functions: Theory and Applications

A Robust Credal Assignment Solution Based on the Generalized Bayes’ Theorem

International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems

This paper proposes a new assignment solution based on the Generalized Bayes’ Theorem (GBT) which... more This paper proposes a new assignment solution based on the Generalized Bayes’ Theorem (GBT) which aims to establish the best matching between two sets of uncertain data. In order to estimate the effectiveness of the proposition, it is compared to the best credal assignment solutions and the well known Global Nearest Neighbor (GNN) algorithm, through synthetic data and a literature example of multi-target tracking scenarios. Given the same input data, the proposed solution gives better assignment results, especially when sensor imprecision increases. However, the proposed solution stills actually computationally more complex than the GNN and the solution proposed by Denoeux et al.

Multi-Target PHD Tracking and Classification Using Imprecise Likelihoods

International Journal of Approximate Reasoning

Abstract This article proposes a method to track and classify multiple target based on kinematics... more Abstract This article proposes a method to track and classify multiple target based on kinematics data. On one hand, tracking is performed using a Probability Hypothesis Density (PHD) filter avoiding the association stage, necessary for many tracking algorithms. On the other hand, Belief Functions and imprecise probabilities are used for the classification task, reducing errors from standard Bayesian classifiers when data are ambiguous. The proposed method is evaluated on several scenarios of multiple aircraft tracking. It is shown in particular that when the number of targets is varying, the proposed approach leads to a reduced number of false created target and improves the classification task over a standard Bayesian classifier where both belief function based classifier and imprecise probabilities classifier give the same result.

Classification crédale multi-cibles Multi-targets evidential classification

This paper deals with the association step in a multi-sensor multitarget tracking process. A new ... more This paper deals with the association step in a multi-sensor multitarget tracking process. A new parameterless credal method for track-to-track assignment is proposed and compared with parameter-dependent methods, namely: the well known Global Nearest Neighbor algorithm (GNN) and a credal method recently proposed by Denoeux et al.

This paper presents a distributed solution for multiobject tracking and classification. The state... more This paper presents a distributed solution for multiobject tracking and classification. The state of objects is partially observed by a set of sensors organized in a network. The idea is to exchange partial data throughout the network and provide a complete information at each sensor level. The proposed solution involves a finite time average consensus where existing solutions are based on asymptotic consensus. The consensus algorithm intervenes in both distributed tracking and classification of multiple objects. It is firstly used to complete information about objects trajectories and secondly to complete beliefs concerning the classification. Simulation results show the relevance of the proposed solution.

This paper compares several assignment algorithms in a multitarget tracking context, namely: the ... more This paper compares several assignment algorithms in a multitarget tracking context, namely: the optimal Global Nearest Neighbor algorithm (GNN) and a few based on belief functions. The robustness of the algorithms are tested in different situations, such as: nearby targets tracking, targets appearances management, etc. It is shown that the algorithms performances are sensitive to some design parameters. It as well shown that, for kinematic data based assignment problem, the credal assignment algorithms do not outperform the standard GNN algorithm.

2013 Workshop on Sensor Data Fusion: Trends, Solutions, Applications (SDF), 2013

Multi-target tracking using multiple sensors is an important research field in application areas ... more Multi-target tracking using multiple sensors is an important research field in application areas of mobile systems and military applications. This paper proposes a decentralized multi-sensor, multi-target tracking and belief (credal) based classification approach, applied to maritime targets. A given number of sensors, considered as unreliable, are designed to locally predict and update targets states using Interacting Multiple Model (IMM) algorithms (one IMM for one target). Targets IMMs are updated by sequentially acquired measurements. The measurements are assigned to the targets by the means of a generalized Global Nearest Neighbor (GNN) algorithm. The generalized GNN algorithm is able to provide information on the newly detected or non-detected targets and these information is used by score functions which manage the targets appearances and disappearances. In addition to the tracking task of multiple targets, each sensor performs a local classification of each one of the targets. The unreliability of the sensors makes the local classifications weak. In this article, a global classification method is shown to improve the sensors classification performances.

Object tracking and credal classification with kinematic data in a multi-target context

Information Fusion, 2014

La redondance d'informations est une solution habituellement proposée pour corriger l'incertitude... more La redondance d'informations est une solution habituellement proposée pour corriger l'incertitude des instruments de mesure observant un système complexe. Dans cet article, on s'intéresse au problème de suivi et de classification de plusieurs cibles à l'aide d'un ensemble de capteurs plus ou moins fiables. Chaque capteur est supposé équipé d'un calculateur lui permettant de suivre et de classer plusieurs cibles effectuant divers mouvements. Le suivi des cibles étant optimalement assuré par des IMM (Interacting Multiple Models) à base de filtres de Kalman, le résultat de la classification reste dépendant des erreurs de mesure. En vue d'avoir le meilleur résultat de classification possible, on s'est proposé de fusionner les classifications locales des capteurs, suivant différentes règles de combinaison des cadres probabiliste et crédal, et de comparer les résultats.