The demand for individualized products is constantly growing. To satisfy this, the production lines need to be built much more flexible. A promising approach to achieve this flexibility, could be a "transport-by-throwing" approach, which is based on the human throwing and catching. The goods, which need to be transported between machines, would be thrown back and forth by these. To ensure a safe and reliable transportation this way, the machine must be able to recognize and catch the transported objects by itself. This task can be roughly divided into three parts: image capturing using two cameras, image processing in order to identify the thrown object and a forecast system to predict the trajectory of the object. Because these tasks are associated with a high computational effort and they have to be calculated in real time (a slow calculation leads to an unsuccessful catching of the object), their implementation on a system-on-chip (SoC) platform, which consists of CPUs and a FPGA, is tested for feasibility. This platform should be able to handle the needed computational effort. In this diploma thesis it is shown, which of the tasks are implemented in Hardware (FPGA) and which in Software (CPU). The results show, that on the used SoC platform, all designed FPGA-parts are able to process the data fast enough, in order to achieve the needed frame rate of 100 frames per second. However the whole system, despite the fact that the FPGA-parts by itself are fast enough, is not able to reach the expected results, because the cameras and the related driver load the two CPUs fully. The achieved frame rate is highly variable between 1 and 25 frames per second, which is far from the expected results. The cameras use a USB 3.0 interface, which is made available through a USB 3.0 expansion card in a PCI-Express slot. Combined with the built-in ARM CPUs, which are clocked at a frequency of 800 MHz, the limits of the used technology has been reached. Implementing a reduction of the data, which is transmitted by the cameras, so that only the relevant part of the picture is sent, the used SoC platform should show better results. However, this approach is, with the currently available camera driver, not optimally usable and leads to errors during the capture. Another possible approach to reach better functionality, would be to port the designed architecture of this diploma thesis to the next generation of the SoC platform, which offers more powerful ARM CPUs.