Netherlands-based unmanned aircraft systems (UAS) maker Aerialtronics says it is taking drone autonomy to the next level by developing a smart dual camera that packs massive onboard processing power with both digital daylight and thermal camera sensors.
The prototype camera, which incorporates an NVIDIA Tegra K1 onboard processor, provides detailed algorithms and makes autonomous devices more intelligent. The project, led by Michele Moscaritolo, head of vision at Aerialtronics, is a giant step forward in onboard data analysis and processing with drones, says the company.
Robin van de Putte, chief of product strategy at Aerialtronics, says, “This is going to be a game changer. Smart payloads will be the next emphasis within the drone industry.“
The device can process both raw data streams from the day-night camera – allowing Aerialtronics’ Altura Zenith UAS to be used more effectively in a multitude of onboard computing tasks.
Aerialtronics says its customers already use aerial data from UAS to inspect, for example, mobile communication towers, electric pylons, bridges and oilfield pipelines. With this camera, the Zenith is able to recognize the detailed shape, type and materials used in these structures and enhance details for 2D and 3D mapping.
“The great thing about working with this [NVIDIA] open chip is that we can inject any algorithm we like, and more importantly, so can the user,” explains Jan Wouter Kruyt, head of research and development at Aerialtronics.
“We have already gone a step further by integrating a daylight and digital thermal camera with NVIDIA’s Tegra K1 chip. No one else offers this powerful data combination. This means the user can analyze both data streams and use the information. Right now, it allows users to do smart things with the images. Eventually, we will be connecting this to the flight computer, and then the whole drone will become smart enough to respond to what it is seeing,” Krut says.
“This is a major step forward in the level of autonomy for drones,” he continues. “Now you can pre-program a drone, and it will fly on a path you set for it. But if you integrate machine vision on an embedded chipset, you create something more intelligent because it can identify what is around it.”