Search

National audience; Nous présentons dans cet article une méthode de détection de personnes et une mesure de leur activité dans des séquences vidéos. En posant l'hypothèse que notre caméra est fixe,nous utilisons tout d'abord la soustraction de l'arrière-plan pour réduire l'espace de recherche de notre classifieur. Nous utilisons ensuite le suivi de points d'intérêt et l'analyse des composantes connectées détectées par la soustraction de l'arrière-plan pour suivre les objets dans le plan image. La classification est réalisée grâce à plusieurs cascades de classifieurs boostés. Cette méthode a été évaluée sur une large base de vidéos. Nous présentons enfin une méthode pour caractériser l'activité des personnes présentes dans une pièce.

International audience

International audience

International audience; We present in this paper a new architecture for remote banking and e-commerce applications. The proposed solution is designed to be low cost and provides some good guarantees of security for a client and his bank issuer. Indeed, the main problem for an issuer is to identify and authenticate one client (a cardholder) using his personal computer through the web when this client wants to access to remote banking services or when he wants to pay on a e-commerce site equipped with 3D-secure payment solution. The proposed solution described in this paper is MasterCard Chip Authentication Program compliant and was experimented in the project called SOPAS. The main contribution of this system consists in the use of a smartcard with a I2C bus that pilots a terminal only equipped with a screen and a keyboard. During the use of services, the user types his PIN code on the keyboard and all the security part of the transaction is performed by the chip of the smartcard. None information of security stays on the personal computer and a dynamic token created by the card is sent to the bank and verified by the front end. We present first the defined methodology and we analyze the main security aspects of the proposed solution.

International audience; We present in this paper a method to analyze the security problems which can occur in a NFC mobile phone payment. In the first section, we give a simplified description of the technical realization for this pilot limited to a part of the system. In the second section, we present the security concepts as a major issue in this scenario. Roughly main requirement is to have same level of security as standard payment EMV transaction. The methodology described in the third section concerns the application payment and its communication with the payment terminal through a NFC link. This study, realized for the ITEA SmartTouch project, starts from a pilot experience that is carried out at Strasbourg in France, in 2007. This trial is the first experiment of a NFC-based payment application that fully supports the international EMV standard and the PayPass program. The end of this article presents different research perspectives to overstepping the encountered problems.

National audience; Nous présentons dans cet article un algorithme de détection et de suivi d'humains, basé sur la vision infrarouge. L'objectif est ici d'avoir des informations fiables sur l'occupation d'une pièce. Nous utilisons pour cela, une segmentation de l'avant-plan avec une modélisation de chaque pixel de l'arrière-plan par une distribution gaussienne, un suivi de cibles basé sur le recouvrement des composantes connectées et une classification basée sur une cascade de classifieurs boostés. Cet algorithme nous permet d'avoir un historique de l'occupation d'une pièce quelle que soit la luminosité. Les résultats expérimentaux montrent l'éfficacité de l'algorithme proposé.

International audience; Many works in the literature focus on the definition of evaluation metrics and criteria that enable to quantify the performance of an image processing algorithm. These evaluation criteria can be used to define new image processing algorithms by optimizing them. In this paper, we propose a general scheme to segment images by a genetic algorithm. The developed method uses an evaluation criterion which quantifies the quality of an image segmentation result. The proposed segmentation method can integrate a local ground truth when it is available in order to set the desired level of precision of the final result. A genetic algorithm is then used in order to determine the best combination of information extracted by the selected criterion. Then, we show that this approach can either be applied for grey-levels or multi-components images in a supervised context or in an unsupervised one. Last, we show the efficiency of the proposed method through some experimental results on several gray-levels and multi-components images.

For ages, humans recognized themselves according to different characteristics (appearance, behavior…). Biometrics is a well known technique to identify an individual or verify its identity; as, for example, fingerprints have been used for more than 100 years to identify one criminal. With computers, this analysis can be realized very quickly and with a higher reliability. Biometrics has many applications: site monitoring (Bird, Masoud, Papanikolopoulos, & Isaacs, 2005), e-commerce (Jain & Pankanti, 2006)… The main benefits of biometrics are to provide better security and to facilitate the authentication process for a user. For example, it can be easy to obtain the password of a user, but it is more difficult to look like the user if a face recognition system is used for the user verification. Biometrics can also provide many advantages for particular applications. Indeed, biometric authentication can be realized in a contactless way that could be important for cultural aspects or reasons of hygiene. For all these motivations, biometrics is an emergent technology that could be more present in our daily life. The goal of this chapter is to make an overview of biometrics. We focus on the authentication process, whose goal is to verify the identity of an individual. Ideal biometric information must have multiple properties: • Universality: all individuals must be characterized by this information; • Uniqueness: this information must as dissimilar as possible for two different persons; • Permanency: it should be present during the whole life of an individual; • Collectability: it can be measured (in a easy way); • Acceptability: it concerns the possibility of a real use by users. The plan of this chapter is given below. The background part presents the different biometric modalities studied in the research labs and used in real conditions. The main thrust of this chapter is an analysis of the benefits and limitations of biometric authentication. We present also the general architecture of a biometric system. Future trends stress the different research topics that should be treated to improve the biometric authentication. It concerns the combination of different biometric systems and their performance evaluation. We conclude by resuming the main aspects of this domain.

This paper proposes a novel Local Gabor Fisher Classifier (LGFC) for face recognition. Gabor feature vector has been recognized as one of the most successful face representations, however, its dimension is too high for fast extraction and accurate classification. In LGFC, Local Feature Analysis (LFA) is exploited to select the most informative Gabor features (hereinafter as local Gabor features) optimally. The selected low-dimensional local Gabor features are then classified by Fisher Linear Discriminant (FLD) for final face identification. We demonstrate that Gabor representation is much more robust than gray-level intensity to image variation caused by the imprecision of facial feature localization. Comparative studies of different similarity measures to local fisher classifier and local Gabor fisher classifier are also performed. The experiments on two traditional face databases, ORL and Aberdeen, have shown that compared with other face recognition schemes, the proposed method can effectively reduce the dimensionality of Gabor features and greatly increase the recognition accuracy.

J'ai obtenu mon Doctorat en 1999 sur le thème de la segmentation d'images. Depuis septembre 2000, je suis maître de conférences à l'ENSI de Bourges. J'enseigne des cours de traitement d'images et d'intelligence artificielle au niveau Master. Je suis depuis le 1er septembre 2006, le responsable des relations européennes et internationales au sein de cet établissement. Je suis rattaché au Laboratoire Vision Robotique (UPRES EA 2078) et à l'équipe Signal, Image et Vision. Je suis le responsable du groupe Interprétation au sein de cette équipe regroupant 3 maîtres de conférences, un ATER et 2 doctorants. Mes activités de recherche principales sont la modélisation 3D d'objets, la détection/reconnaissance d'objets et l'évaluation d'algorithmes de traitement d'images. Ce manuscrit concerne ce dernier thème correspondant à mon activité majeure de recherche. J'ai participé au co-encadrement de 4 doctorants dont deux spécifiquement sur ce thème (les deux ayant soutenu) et deux étudiants de Master recherche. J'ai valorisé mes activités de recherche au travers de plusieurs publications : 1 chapitre de livre international, 6 revues internationales, 2 revues nationales, 53 conférences internationales et 10 conférences nationales. Je suis titulaire du Contrat d'encadrement doctoral et de recherche (PEDR) depuis le 1er octobre 2005. J'ai été sollicité pour effectuer de nombreuses expertises (projets ANVAR, référé d'articles et projets ANR). J'ai participé à de nombreux contrats industriels afin de contribuer au budget de fonctionnement du groupe Interprétation. Ce manuscrit comporte 2 parties. La première correspond à une synthèse de mes activités pédagogiques, administratives et recherche. La seconde présente un résumé des mes activités de recherche dans le domaine de l'évaluation d'algorithmes de traitement d'images. Elle est constituée de trois chapitres. Le premier a pour objectif de positionner la problématique de l'évaluation d'algorithmes de traitement d'images. Les différentes approches utilisées dans la littérature y sont présentées. J'ai développé une méthodologie visant à analyser des approches d'évaluation et à en faciliter leurs conceptions et utilisations. Cette démarche est déclinée au travers de deux algorithmes de traitement d'images. Le troisième chapitre de ce manuscrit présente les travaux sur l'évaluation de résultats de compression d'images ultrasonores dans lesquels j'ai été impliqué. Le thème de l'évaluation de la segmentation d'images que j'ai initié au sein du laboratoire est abordée dans le quatrième chapitre. Je conclus ce document par mon projet de recherche.

International audience; The paper proposes a method for shape recovery of 3D models given a dataset of 3D points on or near the surface of a real object. The developed method is based on genetic programming and provides at the same time the modeling and fitting of a 3D object. The parametrized shapes are defined by the so-called function representation (FRep in short). The final recovered model (available as a FRep) can then be used efficiently in further modeling, animation or analysis. The main advantage of this representation is its low storage cost. We illustrate in the paper some preliminary modeling results of 3D objects.

International audience

International audience; We present in this communication a method, which enables to fit a 3D object defined by a functional representation (FRep) to a dataset of 3D points on its surface. A parametric FRep model sketching the point-set is fitted to the point-set. The best fitted parameters of the model are obtained by using a genetic algorithm, well known for its interesting properties in non-linear optimization. The efficiency of the approach is illustrated for reverse engineering applications.

International audience; This paper describes an industrial system which realizes the cooperation of a robot and a vision system for the quality control of manufactured pieces. We show in this paper its ability to be tele-operated from a distant computer for convenience of development or for a new parameterization of the system. The system we propose is composed of two major parts. The first one concerns the cooperation of the vision and robotic for the control of the quality of an industrial piece. This interaction is controlled by a computer. The second essential element is the ability of this system to be teleoperated from a distant computer to achieve different tasks : definition of some parameters of the quality control (image processing,..),.. We put into obviousness the efficiency of the developed system through some experimental results for the quality control of manufactured pieces.