Menu
Menu
For every vision project, and depending on the problem to be solved, it is necessary to take into account:
For demanding projects in areas where any error could be fatal (aeronautics, military, space etc.), the COTS has difficulty fitting into a global system.
Below are some examples from personal experience along with various projects developed in aeronautics, military or space:
Image pre-processing:
The so-called «industrial» off-the-shelf cameras do not generally have complex image pre-processing on board (e.g. FPN correction, HDR, stabilization). And in complex systems with a very large volume of data to process, it is important to do it as close as possible to the camera head. One can then use an FPGA which plays the role of «Data Cruncher».
Processing power and memory capacity :
When you need to develop an innovative application (e.g. an autonomous vehicle guidance system), you need to use professional sensors with very high performance. High performance means a very large volume of data captured in real time and therefore a high level of computing power is required to process and analyse this data. This can involve running image processing algorithms (for example optical distortion correction, auto exposure) or image analysis algorithms (detection, recognition, identification, tracking…). For example, a PC will be able to process one 12Mpx camera, but if the system consists of six 12Mpx cameras, one PC alone (even very powerful) will not be able to process the generated data flows. It will be necessary to use several in parallel, which will be far from optimal and efficient with bottlenecks sometimes on the computing power side, sometimes on the data flow level.
Form factor :
A COTS camera is already in an imposed format (both concerning the mechanical box but obviously also concerning the electronic boards which compose it). For some projects, the mechanical format of the electronic camera head boards must be specific to meet the size and shape requirements of the housing. For example, in a project where it is necessary to have several camera heads (eg: need to capture in different frequency bands or panoramic vision) it is possible to fit several heads in the same box and share pre-processing if necessary. The same applies to the cooling and management of the optics (mount, motorisation, autofocus…) which can be designed and optimised globally.
Interfaces: compatibility, robustness and bottlenecks :
There will always be compatibility problems between the COTS sub-elements of a global optronics system. These interfaces between elements, real «black boxes», always cost more in the end than custom design. Very often proprietary protocols will be imposed as a real brake in a global vision system. Similarly, certain constraints related to the camera’s working environment (e.g. vibration inside a helicopter) can lead to the creation of custom transmission protocols for high-speed cables. Finally, some proprietary interfaces can constitute real bottlenecks, whereas in a system designed globally, everything is more fluid and optimized. Not to mention the latency in the transmission between the shooting and its processing, which can be crippling in some cases (pilot assistance, for example).
Power consumption and cooling :
The power consumption of an optronics based on COTS cameras will be higher than a custom architecture, even though image pre-processing is not done on board COTS cameras. In a tailor-made system, the need for electronic components is optimised to the bare minimum, thereby reducing consumption and heat emission.
Breakdowns, maintenance and responsibilities :
By thinking about the design of a project in a global way, we eliminate useless maps that interface the different ecosystems, and therefore potential problems of breakdown, problems of finding responsibility and maintenance of the global system. In the end, its reliability is increased.
