Embedded visual sensor systems need to be literally robust not only because of their application domains. The robustness of the methods and algorithms also plays an important role in an environment with frequent interference, e.g. the noise as seen in the grayscale image on top.
Increasingly, smart cameras are used as the camera system. In addition to their photography, these cameras have an internal processing unit for additional image processing. This unit can be a processor, DSPs or FPGAs. Furthermore, these cameras feature data storage and one or more communication interfaces. The added value of this approach is that the camera provides already (pre)processed data for use by the next application layer.
While at the beginning of this century, hardware was an extreme limiting factor, this aspect has meanwhile become secondary. Some consumer devices such as smartphones, which can also be seen as smart cameras, are available cheaply.
Low-cost smart cameras enable the economical realization of various hardware configurations such as visual sensor networks. Here smart cameras serve as agents that require appropriate coordination to achieve the desired results.
Likewise for embedded systems, limited processing speed has been countered by GPU-accelerated parallel processing.
There are many applications of such visual sensor networks, including monitoring of traffic, buildings and processes. A growing share of overall costs falls on energy and software in the development of such distributed optical systems. Especially for embedded systems, which are often harnessed by numerous limitations, demand the choice of the right methods in order to achieve solutions within prescribed budgets and deadlines. SCCH researches and offers solutions in the field of robust embedded visual sensor networks to enable economical and efficient implementations.