This research focuses on the development of field deployed multi-spectral computer vision systems for use on maritime vessels, buoys, ports, and for use on unoccupied aerial systems. The approach includes development of both General Purpose Graphics Processing Unit and multi-core image processing along with Field Programmable Gate Array (FPGA) hybrid architectures for low-power real-time computer vision. The mission is detection and monitoring of security, environmental and safety threats in Arctic maritime environments.
Project Details
This smarter multi-spectral “Go-Pro Like” instrument has the potential for many application areas including safety, security, and resource monitoring. It could be dropped in place or deployed on UAVs or other vessels while consuming a relatively low amount of power (presently 10 Watts or less). In year one we constructed a proof-of-concept system using off-the-shelf hardware components. In year two we bench tested co-processors and found the GP-GPU to be most efficient. Our results have been published and peer-reviewed at the Society for Photographic and Instrumentation Engineering (SPIE) Defense and Commercial Systems (DCS) conference this past spring 2016.
The next steps are to develop power efficient solutions for remote environment security and safety monitoring at the University of Alaska. We plan to install a camera for a year of unattended remote operation in the Alaska environment to assess survivability and operability in Arctic conditions. After this year three test, we plan to pursue expanded use of the SmartCam in the Arctic region beyond security and safety for adjacent scientific uses. Adjacent scientific uses identified include mapping and detection of soil moisture, vegetation, and animal life from a UAV, field tests with military students to see if they can evade detection in a semi-wilderness settings with animal life, and vessel tracking.