In this article we present a generic, flexible and robust approach for an intelligent real-time videosurveillance system. The proposed system is a multi-camera platform that is able to handle different standards of video inputs (composite, IP, IEEE1394). The system implementation is distributed over a scalable computer cluster based on Linux and IP network. Data flows are transmitted between the different modules using multicast technology, video flows are compressed with the MPEG4 standard and the flow control is realized through a TCP-based command network (e.g. for bandwidth occupation control). The design of the architecture is optimized to display, compress, store and playback data and video flows in an efficient way. This platform also integrates advanced video analysis tools, such as motion detection, segmentation, tracking and neural networks modules. The goal of these advanced tools is to provide help to operators by detecting events of interest in visual scenes and store them with appropriate descriptions. This indexation process allows one to rapidly browse through huge amounts of stored surveillance data and play back only interesting sequences. We report here some preliminary results and we show the potential use of such a flexible system in third generation video surveillance system. We illustrate the interest of the system in a real case study, which is the surveillance of a reception desk.

Abstract-This paper builds on an interactive streaming architecture that supports both user feedback interpretation, and temporal juxtaposition of multiple video bitstreams in a single streaming session. As an original contribution, these functionalities can be exploited to offer improved viewing experience, when accessing football content through individual and potentially bandwidth constrained connections. Starting from a conventional broadcasted content, our system automatically splits the native content into non-overlapping and semantically consistent segments. Each segment is then divided into shots, based on conventional view boundary detection. Shots are finally splitted in small clips. These clips support our browsing capabilities during the whole playback in a temporally consistent way. Multiple versions are automatically created to render each clip. Versioning depends on the view type of the initial shot, and typically corresponds to the generation of zoomed in and spatially or temporally subsampled video streams. Clips are encoded independently so that the server can decide on the fly the version to send as a function of the semantic relevance of the segments (in a user-transparent basis, as inferred from video analysis or metadata) and the interactive user requests. Replaying certain game actions is also offered upon request. The streaming is automatically switched to the requested event. Later, the playback is resumed without any offset. The capabilities of our system rely on the H.264/AVC standard. We use soccer videos to validate our framework in subjective experiments showing the feasibility and relevance of our system.

The CANDELA project aims at realizing a system for real-time image processing in traffic and surveillance applications. The system performs segmentation, labels the extracted blobs and tracks their movements in the scene. Performance evaluation of such a system is a major challenge since no standard methods exist and the criteria for evaluation are highly subjective. This paper proposes a performance evaluation approach for video content analysis (VCA) systems and identifies the involved research areas. For these areas we give an overview of the state-of-the-art in performance evaluation and introduce a classification into different semantic levels. The proposed evaluation approach compares the results of the VCA algorithm with a ground-truth (GT) counterpart, which contains the desired results. Both the VCA results and the ground truth comprise description files that are formatted in MPEG-7. The evaluation is required to provide an objective performance measure and a mean to choose between competitive methods. In addition, it enables algorithm developers to measure the progress of their work at the different levels in the design process. From these requirements and the state-of-the-art overview we conclude that standardization is highly desirable for which many research topics still need to be addressed.

In this paper, we propose a new object-based video coding/transmission system using the emerging Motion JPEG 2000 standard [1] for the efficient storage and delivery of video surveillance over low bandwidth channels. Some recent papers deal with JPEG 2000 coding/transmission based on the Region Of Interest (ROI) feature and the multi-layer capability provided by this coding system [2][3]. Those approaches allow delivering more quality for mobile objects (or ROI) than for the background when bandwidth is too narrow for a sufficient video quality. The method proposed here provides the same features while significantly improving the average bitrate/quality ratio of delivered video when cameras are static. We transmit only ROIs of each frame as well as an automatic estimation of the background at a lower frame rate in two separate Motion JPEG 2000 streams. The frames are then reconstructed at the client side without the need of other external data. Our method provides both better video quality and reduced client CPU usage with negligible storage overhead. Video surveillance streams stored on the server are fully compliant with existing Motion JPEG 2000 decoders.

No abstract

This paper tackles the challenge of interactively retrieving visual scenes within surveillance sequences acquired with fixed camera. Contrarily to today's solutions, we assume that no a-priori knowledge is available so that the system must progressively learn the target scenes thanks to interactive labelling of a few frames by the user.The proposed method is based on very low-cost features extraction and integrates relevance feedback, multiple-instance SVM classification and active learning. Each of these 3 steps runs iteratively over the session, and takes advantage of the progressively increasing training set. Repeatable experiments on both simulated and real data demonstrate the efficiency of the approach and show how it allows reaching high retrieval performances.

Video security is becoming more and more important today, as the number of installed cameras can attest. There are many challenging commercial applications to monitor people or vehicle traffic. The work reported here has both research and commercial motivations. Our goals are first to obtain an efficient intelligent system that can meet strong industrial surveillance system requirements and therefore be real-time, distributed, generic and robust. Our second goal is to have a development platform that allows researchers to conceive and easily test new vision algorithms thanks to its modularity and easy setup. This paper focuses on the image analysis modules. It considers the different kind of inputs, algorithm models as well as delay and the need of generality.

No abstract