Person Re-identification Using Spatial Covariance Regions of Human Body Parts

7th IEEE International Conference on Advanced Video and Signal-Based Surveillance, AVSS - 2010 08/2010;
Source: OAI


In many surveillance systems there is a requirement to determine whether a given person of interest has already been observed over a network of cameras. This is the person re-identification problem. The human appearance obtained in one camera is usually different from the ones obtained in another camera. In order to re-identify people the human signature should handle difference in illumination, pose and camera parameters. We propose a new appearance model based on spatial covariance regions extracted from human body parts. The new spatial pyramid scheme is applied to capture the correlation between human body parts in order to obtain a discriminative human signature. The human body parts are automatically detected using Histograms of Oriented Gradients (HOG). The method is evaluated using benchmark video sequences from i-LIDS Multiple-Camera Tracking Scenario data set. The re-identification performance is presented using the cumulative matching characteristic (CMC) curve. Finally, we show that the proposed approach outperforms state of the art methods.

Download full-text


Available from: Sławomir Bąk, Jan 15, 2015
  • Source
    • "(ii) The multi-stage approach with increasing computational cost allows for very efficient re-identification since the computational cheap first stages can be used to reduce the amount of data (candidate models from the database) that has to be considered in the last stage. (iii) Compared to most other state-of-the-art approaches like [3] [4], this approach is applicable in real applications since it is integrated with a detection and tracking strategy. Specifically, it does not rely on manual annotation of people like [3, 4] and builds models for re-identification online without an offline training step like [2] [5]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we approach the task of appearance based person re-identification for scenarios where no biometric features can be used. For that, we build on a person re-identification approach that uses the Implicit Shape Model (ISM) and SIFT features for re-identification. This approach builds identity models of persons during tracking and employs these models for re-identification. We apply this re-identification, which was until now only evaluated in the infrared spectrum, to data acquired in the visible spectrum. Furthermore we evaluate view independence of the re-identification approach and introduce methods that extend view invariance. Specifically, we (i) propose a method for online view-determination of a tracked person, (ii) use the online view-determination to generate view specific identity models of persons which increase model distinctiveness in re-identification, and (iii) introduce a method to convert identity models between views to increase view independence.
    Full-text · Conference Paper · Aug 2011
  • Source
    • "As we see from figure 8, performance of DCD is far behind performance of the other approaches. This is due to the strong environment and camera type induced color differences between the cameras, which are not (see [1]) resolvable by color calibration between cameras. Performance of ISM re-identification is comparable with Haar performance. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we present an approach for person re-identification in multi-camera networks. This approach employs the Implicit Shape Model and SIFT features for person re-identification. One important property of the re-identification approach is that it is closely coupled to a person detection and tracking and uses SIFT feature models which are built during the tracking. We hold this coupling to be an important point because re-identification depends on models that are to be acquired during tracking. These models are then used to re-identify a person when it reappears in the system's field of view. Re-identification itself is performed in a 3-staged approach which allows for efficient re-identification and is perfectly suited for distributed processing where bandwidth concerns are relevant. We show that this re-identification approach which was formerly only evaluated for single camera person re-identification can be successfully applied to the task of multi-camera re-identification. Evaluation in a challenging real-world multi-camera scenario shows that the generic approach which does not use color or other sensor specific features and thus is applicable independently of such sensor specifics shows performance at least comparable to specialized state-of-the-art approaches.
    Full-text · Conference Paper · Jun 2011
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Avec le développement de la vidéo-protection, le nombre de caméras déployées augmente rapidement. Pour exploiter efficacement ces vidéos, il est indispensable de concevoir des outils d'aide à la surveillance qui automatisent au moins partiellement leur analyse. Un des problèmes difficiles est le suivi de personnes dans un grand espace (métro, centre commercial, aéroport, etc.) couvert par un réseau de caméras sans recouvrement. Dans cette thèse nous proposons et expérimentons une nouvelle méthode pour la ré-identification de piétons entre caméras disjointes. Notre technique est fondée sur la détection et l'accumulation de points d'intérêt caractérisés par un descripteur local. D'abord, on propose puis évalue une méthode utilisant les points d'intérêts pour la modélisation de scène, puis la détection d'objets mobiles. Ensuite, la ré-identification des personnes se fait en collectant un ensemble de points d'intérêt durant une fenêtre temporelle, puis en cherchant pour chacun d'eux leur correspondant le plus similaire parmi tous les descripteurs enregistrés précédemment, et stockés dans un KD-tree. Enfin, nous proposons et testons des pistes d'amélioration, en particulier pour la sélection automatique des instants ou des points d'intérêt, afin d'obtenir pour chaque individu un ensemble de points qui soient à la fois les plus variés possibles, et les plus discriminants par rapport aux autres personnes. Les performances de ré-identification de notre algorithme, environ 95% d'identification correcte au premier rang parmi 40 personnes, dépassent l'état de l'art, ainsi que celles obtenues dans nos comparaisons avec d'autres descripteurs (histogramme de couleur, HOG, SIFT).
    Full-text · Article · Jan 2010
Show more