SuMMeR
Report about SuMMer, short for MultiMedia Retrieval, a research project which took place from 2000 - 2002.
From January 2000 until April 2002, a consortium of Dutch organisations participated in the SuMMeR project. In SuMMeR, an acronym for Secure MultiMedia Retrieval, research into new forms of storing and retrieving still and moving images was conducted and two 'Demonstrator' applications were developed to demonstrate the research results. Every organisation brought in specific expertise and goals.
V2_ was one of two 'content partners' involved, bringing in the V2_Archive image collection. Furthermore, V2_Lab actively participated as a technology and design partner, driving the model-view-controller framework approach for the SuMMeR Demonstrators' user interfaces, and designing and building these user interfaces.
DEMONSTRATORS
The V2_ Demonstrator allowed its users to find images in a subselection of the V2_Archive by rating example images or selecting dominant colors or keywords. The query by example mechanism was based on image features; the query by dominant colour was based on a simple image classification.
The AVV Demonstrator allowed its users to find video fragments of highway traffic flow based on highway and hectometer numbers, date and time and type of occurrence (such as a traffic jam). The occurrences were derived through feature extraction and concept derivation.
SYSTEM ARCHITECTURE
The system architecture was set up to be secure and distributed, with three main technology layers, based on open source technology. The Maverick model-view-controller Java/XSLT-framework was used to connect the query and user interface layers. The server was running Linux.
In the database layer, CTIT's MOA (Monet Object Algebra) system was connected to a MySQL database, enabling very fast querying and feature comparisons.
In the query/result layer, queries fired from the user interface were transformed into XQuery format and dispatched to MOA. The results retrieved from MOA were postfiltered using security rules and sent as XML to the user interface.
In the user interface layer, XSLT templates were used to transform the XML resultset into a small set of DHTML pages.
FEATURE EXTRACTION AND CONCEPT DERIVATION
For the V2_ Demonstrator, a set of images from the V2_Archive was processed by a small number of offline feature extractors. These feature extractors stored their features (such as a colour histogram) in the database as vectors. From the features of an image, its dominant color was derived. The query by example functionality compared the features of an example image with all images in the database.
For the AVV Demonstrator, a much more advanced, offline feature extractor was developed that could analyse the video footage to identify the size, speed and location of moving objects on a highway. These features were stored in the database and processed by a concept derivator that was able to identify a number of situations, such as a traffic jam.
SECURITY
A number of security-related features were implemented in the V2_ Demonstrator. Users would connect to the Demonstrator over an SSL-encrypted connection, using a client certificate. This certificate, issued by certificate authority KPN, contained either one or multiple user roles. Depending on the user's role selection, the user interface offered more or less information and functionalities.
Furthermore, a Digital Rights Management system was developed, based on Adobe Acrobat, that limited users in their printing and saving functionality, based on their role.
The results of the SuMMeR project were presented during a workshop at KPN Research headquarters in Leidschendam.
