Logo multitel

Organized in the framework of portfolio; and platforms.

See http://mediatic.multitel.be/ for more details.

4th Multitel Spring workshop on video analysis – June 02, 2009

  • 10:00 Welcome and coffee
  • 10:25 Opening Remarks and Welcome

Paper Session 1: Surveillance & applications

  • 10:30 “Scenario recognition in video-surveillance by combining probabilistic graphical approaches”, Cina Motamed, University of Littoral Cote d’opale, LASL Laboratory.

In this work we investigate the recognition of human behavior from video sequences by using an active perception strategy. The scenarios recognition algorithm is based on a combination of probabilistic graphical techniques. The model of scenario contains two main layers. The first one allows highlighting low level events from the observed visual features as the object trajectory. The second layer is focused on the temporal reasoning stage by using Bayesian Networks. The system mainly takes the advantage of an event based approach as a flexible temporal reasoning capability and it uses conventional HMM approach for the trajectory recognition.

  • 11:00 “Ground-Target Tracking in Multiple Cameras Using Collaborative Particle Filters and Principal Axis-Based Integration”, ULG, Wei Du.

This paper presents a novel approach to tracking ground targets in multiple cameras. A target is tracked not only in each camera but also in the ground plane by individual particle filters. These particle filters collaborate in two different ways. First, the particle filters in each camera pass messages to those in the ground plane where the multi-camera information is integrated by intersecting the targets’ principal axes. This largely relaxes the dependence on precise foot positions when mapping targets from images to the ground plane using homographies. Second, the fusion results in the ground plane are then incorporated by each camera as boosted proposal functions. A mixture proposal function is composed for each tracker in a camera by combining an independent transition kernel and the boosted proposal function. The general framework of our approach allows us to track individual targets distributively and independently, which is of potential use in case that we are only interested in the trajectories of a few key targets and that we cannot track all the targets in the scene simultaneously.

  • 11:30 “Multi-feature stereo vision system for road traffic analysis”, Jacek Czyz, Macq Electronic, Juan Carlos Tocino, ULB, Quentin Houben, ULB.

This paper presents a method for counting and classifying vehicles on motorway. The system is based on a multi-camera system fixed over the road. Different features (maximum phase congruency and edges) are detected on the two images and matched together with local matching algorithm. The resulting 3D points cloud is processed by maximum spanning tree clustering algorithm to group the points into vehicle objects. Bounding boxes are defined for each detected object, giving an approximation of the vehicles 3D sizes. A complementary 2D quadrilateral detector has been developed to enhance the probability of matching features on vehicle exhibiting little texture such as long vehicles. The algorithm presented here was validated manually and gives 90% of good detection accuracy.

  • 12:00 “Intelligent video-surveillance solutions: Panorama Activity Detection case study”, Jean-François Delaigle, Christophe Parisot, ACIC.

In this work, we present the use of Pan-Tilt cameras for the surveillance of large areas, e.g. the see. The presentation describes the main steps of the algorithm as well as the technical choices for the system components in a real deployment.

Lunch break

(A buffet lunch will be offered by Multitel)

Paper Session 2: Human body modelling

  • 13:30 “Monocular human upper body pose estimation for sign language analysis”, Nicolas Burrus, ULG.

We present a system to track human upper body using a single camera. The goal is to extract relevant features for sign language recognition, such as location, velocity and configuration of forearms and hands. Motion blur, rapid moves, self-occlusions and non-rigid deformations make the independent tracking of individual part difficult, ambiguous and not very reliable. Thus, we experiment top-down approaches based on pictorial models which aim at simultaneously modeling the geometry of human parts, the appearance of each part and the temporal continuity in a unified statistical framework. First results will be shown on the NGT corpus of Dutch sign language videos.

  • 14:00 “Detection and analysis of people in APIDIS framework”, Damien Delannay, Fan Chen and Christophe De Vleeschouwer, UCL.

This talk surveys some of the initial results of the FP7 APIDIS project, dedicated to the autonomous production of images based on intelligent sensing. The methods presented during the talk aim at detecting and recognizing players on a sport-field, based on a distributed set of loosely synchronized cameras. Detection assumes player verticality, and sums the cumulative projection of the multiple views foreground activity masks on a set of planes that are parallel to the ground plane. After summation, large projection values indicate the positions of the players on the ground plane.

Paper Session 3: Computer vision frameworks & interfaces

  • 14:30 “Automomous production of sport video sequences in APIDIS framework”, Fan Chen, Damien Delannay, and Christophe De Vleeschouwer, UCL.

This talk surveys some of the initial results of the FP7 APIDIS project, dedicated to the autonomous production of images based on intelligent sensing. Given the positions of the ball and players, we propose an autonomous system for personalized production of sport event reports. The purpose is to render the entire action, while rendering each individual object with sufficient resolution. We propose criteria for optimal planning of viewpoint coverage and camera selection for improved story-telling and perceptual comfort. By using statistical inference, we design and implement the estimation process. Experiments are made to verify the system, which shows that our method efficiently alleviates flickering visual artifacts due to viewpoint switching, and discontinuous story-telling artifacts.

  • 15:00 “Computational attention in video processing” , Matei Mancas, FPMS.

For the Numediart project (www.numediart.org) the modelisation of human attention for video signals was applied to gestures expressivity and emotion caracterisation. The difference between motion attention and motion detection is explained and some applications in videosurveillance scenarios will be proposed.

  • 15:30 “Learning approach for multicontent analysis of compound images”, Quentin Besnehard, Cedric Marchessoux, Tom Kimpe, BARCO N.V. (Belgium).

In the context of the European Cantata project (ITEA project, 2006-2009), within Barco, a complete Multi-Content Analysis framework was developed for detection and analysis of compound images. The framework consists of: a dataset, a Multi-Content Analysis (MCA) algorithm based on learning approaches, a Ground Truth, an evaluation module based on metrics and a presentation module. The aim of the MCA methodology presented here is to classify image content of computer screenshots into different categories such as: text; Graphical User Interface; Medical images and other complex images. The AdaBoost meta-algorithm was chosen, implemented and optimized for the classification method as it fitted the constraints (real-time and precision). A large dataset separated in training and testing subsets and their ground truth (with ViPER metadata format) was both collected and generated for the four different categories. The outcome of the MCA is a cascade of strong classifiers trained and tested on the different subsets. The obtained framework and its optimization (binary search, pre-computing of the features, pre-sorting) allow to re-train the classifiers as much as needed. The preliminary results are quite encouraging with a low false positive rate and close true positive rate in comparison with expectations. The re-injection of false negative examples from new testing subsets in the training phase resulted in better performances of the MCA.

  • 16:00 “Progressive learning for Multiple View 3D Object Retrieval”, Mathieu Coterot.

In this talk, we consider the challenge of retrieve object in a 3D database. Our previous work was about an interactive system able to retrieve scenes in videosurveillance data, based on SVM classification, relevance feddback and active learning methods. Starting to this point, we look for adapt this to retrieve 3D objects. Using the ‘viewer-centered representation scheme’, our system consider 3D objects as a set of 2D views. Objects descriptions is thus reduced to a serie of 2D features, easily extractable and comparable. Contrary to a 3D descritpions, this method offer the possibility to search a 3D Object by only one (or several) of its 2D projections.

  • 16:45 Discussion & Closing