A hybrid system on chip solution for the detection and labeling of moving objects in video streams

A a system on-chip architecture performing the task of moving object detection and labeling in video data is proposed in the paper. The solution involves task partitioning between the microprocessor-bound software and the dedicated hardware coprocessor. A detailed description of the system is given, along with the analysis of the hardware/software decomposition process. A summary of the resources used for the implementation and the projections of system performance are also given in the paper. The resulting solution can be applied either as a standalone smart camera solution or a part of a bigger system.