S.B. Borbély, C. Zinner, and W. Kubinger (Austria)
Modeling and Simulation Methods, Simulation Tools and Techniques, Embedded Systems, Computer Vision.
This paper describes a proposed software design process with model-based evaluation approach to high performance image processing algorithms for Embedded Vision Systems (EVS). The resulted library (so called PfeLib) encloses software primitives which are optimized target specific to reach a high execution performance. However, due to the high computational complexity of most computer vision algorithms, realization of such applications on embedded systems is not a straight forward task. Therefore, the functional behavior of each algorithm was evaluated by the use of MATLAB Simulink. The main goal was to accelerate development and to improve the quality of the resulting algorithms and applications by sys tematic design, development and test. Target specific code generation was performed after modeling and simulation, where the generated code was evaluated and compared to the hand optimized code of the PfeLib. This resulted in Software In the Loop (SIL) simulation of the hand coded algorithms using MATLAB Simulink. Two Texas Instru ments DSP Starter Kit (DSK) boards were used for eval uation and In-the-Loop testing. For instrumentation of the generated code, the used CPU cycles of the core part of the algorithm are counted as a measure for the mean execution time.
Important Links:
Go Back
