Dana Codreanu

A distributed architecture for flexible multimedia management and retrieval

Developing an efficient system that manages distributed multimedia content supposes to minimize resource consumption while providing the most relevant results for a users query in the shortest time. This paper presents LINDO, a generic architecture framework for distributed systems that acquires efficiency in multimedia indexing and retrieval. Three characteristics particularize it: (1) it differentiates between implicit algorithms executed over all the multimedia content at the acquisition time, and explicit algorithms, executed on demand for answering a specific need; (2) it stores and processes multimedia content and metadata locally, instead of transferring and indexing it on a central server; (3) it selects a set of relevant servers for query execution based on the user query semantic processing and on the system knowledge, including descriptions of distributed servers, multimedia content and indexing algorithms. The paper relies on a concrete implementation of the LINDO framework in order to validate this contribution.

Video Spatio-Temporal Filtering Based on Cameras and Target Objects Trajectories -- Videosurveillance Forensic Framework

This paper presents our work about assisting video-surveillance agents in the search for particular video scenes of interest in transit network. This work has been developed based on requirements defined within different projects with the French National Police in a forensic goal. The video-surveillance agent inputs a query in the form of a hybrid trajectory (date, time, locations expressed with regards to different reference systems) and potentially some visual descriptions of the scene. The query processing starts with the interpretation of the hybrid trajectory and continues with a selection of a set of cameras likely to have filmed the spatial trajectory. The main contributions of this paper are: (1) a definition of the hybrid trajectory query concept, trajectory that is constituted of geometrical and symbolic segments represented with regards to different reference systems (e.g., Geodesic system, road network), (2) a spatio-temporal filtering framework based on a spatio-temporal modeling of the transit network and associated cameras.

Adapting Indexation to the Content, Context and Queries Characteristics in Distributed Multimedia Systems

Due to the dramatically increasing amount of multimedia contents in several application domains, effective and flexible solutions for distributed data indexation are essential. In the management of these multimedia contents, the indexing process is the most important resource consumer, in terms of data transfer over the network and CPU consumption. As result of our research, we located two points for reducing resource consumption: the limitation of multimedia content transfer over the network for indexing, as well as the reduction of multimedia indexing amount by employing only the most appropriate algorithms for performing indexation, only over the relevant contents. We present in this paper the solution developed in the context of the LINDO project for indexing multimedia content into a distributed system that originally address these two points by reducing as much as possible the resource consumption through (1) a distributed indexing technique that avoids multimedia transfer, (2) a flexible mechanism for selecting the indexing algorithms to be employed on each remote server, according to the multimedia content characteristics, its acquisition context and user queries history.

Modelling Multimedia Indexing Algorithms Description for Implicit and Explicit Indexation

The increasing amount of multimedia data is generally indexed by algorithms which extract metadata that will be further queried by users for retrieving some desired information. Because of the heterogeneity and the diversity of these indexing algorithms, it is not suitable to execute at once all indexing algorithms. Actually, it will overload the information system and will also produce metadata that may not be used inside the system. For example, in a car park surveillance system, it is not necessary to apply to all the video sequences algorithms that detect persons. However, at a certain time someone might want to know about the persons that crossed the parking. It can happen that the multimedia information system returns an empty result to a query because the relevant algorithm for treating the content has not been identified or located. In this case, an algorithm application chain could produce the metadata that could give an answer to the query. In order to identify the right indexing tool(s) for the right resource(s), it would be useful to have a description attached to these tools. In this paper, we propose an indexing algorithm description model and show how to use such a model for determining an appropriate set of indexing algorithms (or sequences of algorithms) according to specific user needs, properties and execution contexts. Our proposal has been integrated in the LINDO project for implicitly and explicitly indexing multimedia contents.

Spatio-temporal metadata filtering and synchronising in video surveillance

This paper presents an ongoing work that aims at assisting video-protection agents in the search for particular video scenes of interest in transit network. The video-protection agent inputs a query in the form of date, time, location and a visual description of the scene. The query processing starts by selecting a set of cameras likely to have filmed the scene followed by an analysis of the video content obtained from these cameras. The main contribution of this paper is the innovative framework that is composed of: 1) a spatiotemporal filtering method based on a spatio-temporal modelling of the transit network and associated cameras, and 2) a content-based retrieval based method on visual features. The presented filtering framework is to be tested on real data acquired within a French National project in partnership with the French Interior Ministry and the French National Police. The project aims at setting up public demonstrators that will be used by researchers and commercials from the video-protection community.

Towards a Pervasive Access Control within Video Surveillance Systems

This paper addresses two emerging challenges that multimedia distributed systems have to deal with: the user’s constant mobility and the information’s sensitivity. The systems have to adapt, in real time, to the user’s context and situation in order to provide him with relevant results without breaking the security and privacy policies. Distributed multimedia systems, such as the oneproposed by the LINDO project, do not generally consider both issues. In this paper, we apply an access control layer on top of the LINDO architecture that takes into consideration the user’s context and situation and recommends alternative resources to the user when he is facing an important situation. The proposed solution was implemented and tested in a video surveillance use case.