Huld is developing its own platform for vision-based navigation with an aim to minimize reliance on GNSS (e.g., GPS, Galileo) services for autonomous systems like unmanned aerial vehicles (UAVs).
Huld wants to address the shortcomings of GNSS services, like GPS, by providing a vision-based system that can geo-reference (search, overly and align) images of Earth taken at lower altitude (< 150 m). Huld accomplishes this task using a single camera mounted on an unmanned aerial vehicle (UAV).
GNSS sensors are an effective method to navigate and localize around the globe. Our devices e.g., phones use them daily. However, to achieve autonomy in an open environment, it is adamant to have better precision, for which GNSS systems are unreliable.
In recent years, high-quality satellite images (maps) have become readily available, allowing image processing algorithms to be more creative. Huld utilizes these satellite image repositories to geo-reference images taken via UAVs (drones). Thus, identifying the correct location of each pixel in real time.
The platform has been trained and tested such that it is impervious to seasonal changes on the Earth’s surface and altitude of the camera providing images. Traditionally similar systems require a dual or 3D camera systems. However, our system is performing all these functions with a single camera.
“A picture is worth a thousand words”, A vision-based system is congruent to this statement. As we geo-reference an image, we open the door for building other applications on-top, like object detection and tracking. Preliminary tests of our platform allowed us to successfully track vehicles on a highway through a single-camera UAV video feed.
Combining the capability of geo-referencing and object detection, the market for this platform includes but not limited to:
We are working with close partners to establish a cloud-based system that takes in-flight footages of UAVs and provides geo-referenced output as return. We can setup a demo for any interested party.