Objective
The principal aim of DETECT is to develop a new generation video analysis and recognition platform, which provides the possibility for semantic block detection (temporal) and ROI detection (spatial), based on sample references from a database and analysis-modules which apply matching and recognition algorithms based on known references (shapes, pattern) on the input video stream. These detected regions are then being subject to further analysis and processing. The project will focus on the applications problem of detecting and delimiting predefined static and dynamic objects. This issue has currently a very large demand for both cultural and economic reasons. Thus one result of DETECT shall be an exploitable application, which provides detailed statistics and ratings about (predefined) commercials and trade-marks within the analysed input stream.
Objectives:
The principle goal of the DETECT project is to implement a general platform for real time detection of semantic blocks and regions within digital video streams. These detected regions are then be subject to further analysis and processing. The project will focus on the application problem of detecting and delimiting predefined static and dynamic objects. This issue has currently a very large demand for both cultural and economic reasons. DETECT is an industrial driven project, although its nature is R&D. The project will result in a prototype, which can be turned into a product after the project. For this reason the main modules are implemented as sample applications (Processing Units) for the categories which are of high commercial interest (e.g.: identification of race-cars and soccer-players).
Work description:
DETECT provides a general platform for a real time analysis for streaming video input and supports three different types of Processing Units (modules) which are as follows:
* Detection of semantic blocks. A semantic block covers the temporal domain only. Thus a typical semantic block-detector just indicates, whether the streamed content is of a certain type or not. Within DETECT the semantic block-concept will be implemented for commercial/advertising blocks, as they appear frequently in television-broadcast. Depending on the outcome of the semantic block detection, a specific detection of regions of interest (further on ROI) can be applied;
* Detection of ROI in order to identify static and dynamic objects. ROI is a specific local region which is due to the nature of the streaming input also related to the temporal domain. Such ROI, whenever identified in real time can be used to restrict further analysis or to simply recognize predefined objects, which match with the ROI. Within DETECT the ROI-concept will be applied to sports-applications (soccer and formula 1) and therein to locally moving objects like soccer-players and race-cars, but also to static objects like hoardings. Each detected ROI can be further analyzed with pattern recognition tools depending on the type of ROI:
* Motion picture analysis. The main objective herein is, to detect predefined (company-) logos stored in a central reference database. Those logos can be trademarks like in the DETECT sample application, but could also be of nearly any other type. As the size of the reference -system has to be scaleable, the logo-detection analysis will be done off-line not as a real time application.
Milestones:
(I) Overall Specification - PM 6: user requirements, user groups & use cases, test scenarios, collection of test material/content, overall system architecture(II) DETECT V0.1 - PM 12: first versions of semantic block detection/ROI/motion analysis modules(III) DETECT V0.5 - PM 21: final versions of semantic block detection/ROI/motion analysis modules(IV) DETECT V1.0 - PM 24: fully integrated DETECT prototype system
Fields of science (EuroSciVoc)
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
- natural sciencescomputer and information sciencesdatabases
- natural sciencescomputer and information sciencesartificial intelligencecomputer visionmotion analysis
- humanitiesartsmodern and contemporary artcinematography
- natural sciencescomputer and information sciencesartificial intelligencepattern recognition
You need to log in or register to use this function
Call for proposal
Data not availableFunding Scheme
CSC - Cost-sharing contractsCoordinator
8010 GRAZ
Austria