Unmanned Systems Technology 038 l Skyeton Raybird-3 l Data storage l Sea-Kit X-Class USV l USVs insight l Spectronik PEM fuel cells l Blue White Robotics UVIO l Antennas l AUVSI Xponential Virtual 2021 report

95 “After adding dual-channel processing in software release 3.2, the 4000-OEM now supports image blending, further reducing the processing function gap with our 3000-OEM,” said Dr Steve Olson. “And our team is working to ensure that the 4000-OEM provides even greater functionality in our 3.4 software update.” The 4000-OEM has dimensions of 50.5 x 38 mm, weighs 13 g and consumes 5 W. It can take multiple digital video inputs, using parallel digital inputs compatible with any SightLine 3000-OEM input adapter board, MIPI and USB-3. It also supports the Sony ER8550 4K and 4K HDMI inputs, using COTS adapter boards. Video can be output over Ethernet with H.264 or H.265 compression, over HDMI and HD-SDI if an HDMI-SDI adapter board is used. The system’s detection capability has also been improved, and it can now support the filtering of objects by classification. Custom classifier tools are available to support customer-owned libraries of images, including support for programming in Python 3.5. A dynamic dead-pixel filter was recently added to the suite of cooled-detector raw video processing functions, allowing users to remove bad pixels in real time. “Offering that on our video processors allows integrators to use them in lieu of a dedicated DPR/NUC card,” Dr Olson added. “Also, taking advantage of our autofocus, enhancement and encoding options provides system flexibility within the same low-SWaP package.” Intelligent payload systems developer Overwatch Imaging told us about its new PT/N Oceanwatch multi- sensor system, which has been designed for automated wide-area maritime search and detection applications using a combination of AI and IR for accurate operations in the day or at night. Colleen Jones said, “The system is designed for fully unconstrained pitch rotation. A downward, more focused search allows for more pixels on target, which is ideal for finding lost small objects of interest such as a person, whereas a more outward scan for maritime patrol lets the operator cover larger areas for bigger objects such as boats.” The PT/N Oceanwatch uses a mixture of proprietary classical and machine learning algorithms to detect non-water objects of interest. Edge-based AI analytics are used to reduce data densities down to actionable intelligence data at the source, with all data processing performed on the payload in real time, significantly compressing file sizes. This enables information to be more easily and cost-effectively downlinked to the data’s customers, whether they are on a ship nearby or elsewhere in the world. The first commercial units were shipped earlier this year to customers in Central America and Europe. The company also discussed its new TK- Series of payloads for automated mapping and object identification for oil spills. As Jones explained, “It can take 4-6 hours after flying over an oil spill to get a map to first responders. An oil-spill removal organisation asked if we could create a combined hardware-software solution to create maps of ‘actionable oil’ in real time.” Actionable oil constitutes the thickest areas of oil in spills, which can typically be collected, dispersed or burnt as part of the clean-up process. The AI and sensors in the payload delineate between actionable oil and sheen (as well as other false positives). Multi-band mosaic maps of actionable oil are generated in real time, with edge AI used as in the PT/N for data compression and easier downlinking. It also identifies other pollutants such as cargo containers and plastics, and can be used for identifying and counting different marine wildlife species (as oil-spill responders will often take wildlife protection into account during their operations). Embention discussed the latest update for its Veronte Autopilot 4x product, which has been carried out to enhance safety for urban air taxi-type vehicles. Javier Espuch said, “To make up for the lack of components optimised for the eVTOL sector, we’ve enhanced the Veronte Autopilot 4x to meet the FAA and EASA’s DO178C and DO254 standards, making it the first quadruple-redundant eVTOL flight controller.” DO178C is the airworthiness certification standard for aviation equipment software, while DO254 is the equivalent for aircraft electronics hardware. Both include development guidelines for engineers to follow for design assurance. “The new eVTOL systems are expected to be manned/autonomous, and most of them will be flown in urban and similarly critical areas,” Espuch added. “Conventional autopilot systems for manned aircraft cannot control these distributed-lift vehicle architectures, and typical UAV autopilot systems are either not compatible with having a pilot on board or cannot provide the reliability evidence needed to achieve certification for use on board eVTOL craft.” The Veronte Autopilot 4x weighs about 700 g, and provides functions including fully autonomous, manual and fly-by-wire control modes, as well as the ability Unmanned Systems Technology | June/July 2021 The Overwatch PT/N multi-sensor system