Implementation of hardware-optimized 3-D wave digital filters for motion-based object detection in video scenes

Even though filtering objects in video scenes using linear filters has been an active research field for many years, the presented results have been more focused on theory than on application, which is mainly due to the high computational complexity of the developed filter structures that former computer systems could not process in real-time. Particularly due to high product quantities and high competition, computation costs of applications like object detection in driver assistance systems, for instance, still have to be as small as possible. This work is focused on 3-D linear shift invariant (LSI) filters targeting the separation of signals due to different directional components. We have implemented wave digital filters with low complexity and therefore low hardware requirements with the aim of developing a cost efficient real-time object detection system. The developed filters have been both simulated using Simulink Fixed Point and implemented on a Xilinx Spartan 3E FPGA Board. Experiments have shown promising results underlining the potential of motion-based object detection to lead to new and effective methods in driver assistance system development.