In order to accommodate the transition from standard to high-definition imagery in unmanned airborne platforms, it is necessary to find a way to reduce the increased transmission bandwidth of the system. One viable solution is to provide an onboard image processing computer that tracks objects of interest and transmits only that portion of the image in high-definition. However, UAV onboard processing must fall within strict SWAP constraints. In order to meet this challenge, multiple mobile processors will be evaluated to determine which has the best balance of processing power, flexibility, ease of programmability and relevant built-in image processing capabilities. The industry standard Qseven carrier board will provide a common platform for testing multiple mobile ARM-based processors. Each processor will be first tested against multiple image processing algorithms, benchmarked for speed and throughput. Ease of programmability will be evaluated as time spent adapting a functioning x86 implementation of each algorithm over two steps, basic functionality of the algorithm and optimized algorithm using features built into the hardware such as a GPU. The results of these tests will then be evaluated to determine if there is a platform that provides the necessary processing power to perform on-board image processing for a UAV.
Benefit: The key benefit is to preserve the ability to stream video offboard UAVs while reducing the size, weight, and power required by the harware. BRE has identified several opportunities to commercialize the technology developed under the proposed effort: 1. Near-term disaster assessment from UAV or Balloon 2. Robotic Video Applications 3. Weather-balloon borne imaging of crop yields
