The project “Parallel Face Recognition in Video Streams” (PaGeVi) is aimed at further developing a method to identify persons wanted at large events. Violent fans, for example, re-present a big security problem at soccer stadiums. By a parallelization of the software, scientists of KIT and the Research Center for Information Technology (FZI) want to accelerate image processing and improve recognizability. Among the application partners are the b.i.g. security service company and the Karlsruher SC (KSC soccer club).
“If we succeed to catch soccer fans who have already been identified as violent at the entrance of the stadium with our method, security would be enhanced considerably,” says Professor Rainer Stiefelhagen from the KIT Institute of Anthropomatics. For tests and further development, the scientists will record data during three home matches of the KSC soccer club at the Wildpark stadium, for the first time on July 31. Staff members of the project partners will act as test persons, normal stadium operation and the remaining visitors will not be affected. It is the objective to detect these test persons automatically via cameras in the crowd at the entrance for guest supporters and near the staircase leading to the area of guest supporters.
For the analysis of the video image sequences, the method has to be run in real time. So far, this requirement has been at the expense of the recognition performance. “Processing speed can only be increased by a parallelization of the software,” says Stiefelhagen. “The question now is: How can a process be divided into partial problems that can be executed in parallel for the best possible results to be achieved at the computing power available?” The basis are manycore systems, e.g. computer systems with several processor cores on a single chip. On these systems, individual images of videos can be processed in parallel, such that also large-expenditure computations can be carried out in real time. Among others, comparison with the persons wanted is a time-critical step.
Apart from the increase in processing speed, parallelization also opens up new possibilities of face recognition. On many-core systems, various face recognition methods can be executed simultaneously. The results of the individual processes can then be combined in order to increase the recognition performance.
Suitability of this method in practice will be tested at the Karlsruhe Wildpark stadium. The project partner b.i.g. gebäude service gmbh will be responsible for the security and monitoring of the facility and will supervise practical evaluation. After the project, the project partners GPP communication and Videmo Intelligente Videoanalyse will implement the process as a marketable software solution.
The project is funded with a total amount of EUR 1.2 million by the Federal Ministry of Education and Research (BMBF) under the KMU-innovativ program.
More information on the program:
(http://www.bmbf.de/de/10785.php)
In close partnership with society, KIT develops solutions for urgent challenges – from climate change, energy transition and sustainable use of natural resources to artificial intelligence, sovereignty and an aging population. As The University in the Helmholtz Association, KIT unites scientific excellence from insight to application-driven research under one roof – and is thus in a unique position to drive this transformation. As a University of Excellence, KIT offers its more than 10,000 employees and 22,800 students outstanding opportunities to shape a sustainable and resilient future. KIT – Science for Impact.