Embedded Vision
Embedded Vision focuses on bringing the power of traditional vision systems into areas where the Size, Weight, and Power (SWaP) of these systems would normally forbid it. This includes applications like self-driving cars, security drones, and facial recognition.
Camera sensor technology "grows" increasingly small, pushed in part by the smartphone market. This has allowed for smaller and smaller cameras, with lower SWaP. Image acquisition is only half the story; traditional image processing still violates SWaP.
Traditional processing occurs on a Central Processing Unit (CPU). This is a device that has to be able to do everything from playing music to surfing the web. Because it has to be capable of everything, it isn't optimized for a specific task.
Graphical Processing Units (GPUs) are better suited to image processing. These components allow massive speed increases compared to CPUs. These may often come into play in the embedded vision space. GPUs are designed for processing large amounts of data in parallel.
While GPUs are great for image processing, they still are not tailor-designed for a specific imaging application. An Application-Specific Integrated Circuit (ASIC) will give you the best power consumption and speed for your design. The downside? Because ASICs are not Commercial Off The Shelf components, they have a very large Non-Recurring Engineering cost. Typically, this approach is only utilized by large OEMs for products in mass production.
There is an alternative: Field-Programmable Gate Arrays (FPGAs). FPGAs are unique in that they allow a custom-tailored circuit to be "written" to the device multiple times. Because of this, FPGA designs can be customized to a specific design while still being a COTS component.
Pyramid Imaging works with its partners to bring all of these solutions to the table. We also have rapid FPGA development capabilities, allowing your vision design to have low SWaP while still being as robust as a desktop application.
For more information about how Pyramid Imaging can help your embedded vision application, please contact us using this form.