UV Online 
Vision processing promises to squeeze more intel from imagery and save bandwidth
As more customers demand very high resolution imagery, including HD, from a rapidly expanding number and variety of imaging sensors, ever smarter image processing will be essential if analysts and warfighters are to make sense of it all in time to act on the intelligence it yields. Just as critically, more of that processing capability will have to reside in the sensors themselves to avoid overwhelming bandwidth-limited communications networks with a deluge of raw imagery.
It is against this background that video processing house Sarnoff Corporation is demonstrating its Acadia II vision processor here at the SPIE Defense, Security and Sensing event in Orlando, Florida.
Sarnoff developed its Acadia line with DARPA support under the MANTIS programme. Acadia II is the latest variant, offering: ‘even better performance and features including real-time video enhancement, stabilisation, mosaicking, multi-sensor video fusion, stereo range estimation, and image feature detection’.
The company describes it as a System-on-Chip (SoC) with the core video processing capabilities hard-wired into an Application Specific Integrated Circuit (ASIC) on the same silicon chip as fully programmable ARM Quad Core processors that can act as the CPU for the systems Acadia II supports.
This capability could trigger fundamental change in the architecture of sensor systems. ‘With more processing power, smaller size, and less energy consumption, don’t think of it as putting Acadia II in your system. The approach is to put your system on Acadia II’, says Mark Clifton, vice-president, products and services at Sarnoff.
Among the ‘next-generation’ applications at which Sarnoff is aiming Acadia II are portable and wearable vision systems; security and surveillance platforms; manned and unmanned aerial and ground vehicles; small arms; border and perimeter protection; and vision-aided GPS-denied navigation.
With massive processing power and, presumably, well-developed algorithms, SoC solutions such as Acadia II now offer the capability to extract actionable intelligence very rapidly within, for example, a sensor on a UAV and transmit that instead of full motion video to the ground. The implied bandwidth benefits are obvious, but it will take thorough comparative trials to convince sceptical end users that, in an increasing range of circumstances, they can do without full motion video.
By Peter Donaldson, Orlando, Florida
Email this to a friend.