Autonomous vehicles are staking their claim to roadways across the world, but the companies developing products for this industry still face hurdles in customer adoption. They need to assure drivers, passengers, and others sharing the road, that their systems can respond in an instant and take the right actions when necessary.

Multi-camera systems paired with high-speed networking technology can make that a reality, though the processing requirements could overload even some of the most capable CPUs on the market. Leveraging the parallel processing enabled by FPGAs system, on the other hand, frees up a centralized CPU from much of that responsibility.

And that’s just what DornerWorks guided a customer to when they were looking for a solution to quickly and efficiently capturing and processing video feeds from multiple sources.

Our Solution

The solution DornerWorks implemented used a pair of Xilinx Zynq UltraScale+ devices was to process up to seven 12MP 12-bit video streams, and up to three 1MP 8-bit video streams simultaneously.

  • All video streams ran at 30 fps and was received via MIPI CSI-2 and parallel interfaces.
  • The video was filtered, color corrected, color space converted, and multiplexed over a PCIe 3.0 x8 link to the system for additional video processing.
  • The un-processed video was sent via a PCIe 3.0 x4 link to the system for logging.