OpenFAS

Leaders:  Prof. Dr.-Ing. Holger Blume
Email:  3621
Team:  Dipl.-Ing. Christopher Bartels
Year:  2012
Sponsors:  "Zentrales Innovationsprogramm Mittelstand" des Bundesministeriums für Wirtschaft und Technologie (BMWi)
Lifespan:  Juni 2012 - Oktober 2013
Is Finished:  yes

The goal of this project is to close the gap between proprietary algorithms research for video-based DAS systems and efficient mapping of these algorithms on product- enabled embedded systems. A methodology is shown and demonstrated on a real example of how a standardized intermediate layer, in the form of an application-specific, optimized open image processing library, between FAS applications and a cost, performance and power dissipation suitable implementation platform ( videantis multicore video processor platform with modules HW acceleration), enables the implementation of proprietary driver assistence algorithms. Algorithms for camera-based DAS applications are to be analyzed, with focus on suitability for real-time implementability on the basis of an accelerated OpenCV image processing library first. A detailed selection of algorithms shall be created. This is followed by a port of the PC-based implementations on accelerated hardware and software modules. Finally, methods and routines are examined with regard to their real- time capability, and optimized. This is demonstrated on the videantis Vantage processor.