Researchers at the American University of Perdue have developed an AI that could well bury current night vision technologies. Called HADAR, this tool relies on thermal detection, infrared imaging and AI to drastically improve visibility at night.
Since the launch of ChatGPT at the end of 2022, artificial intelligence has been on everyone’s lips. This is undoubtedly the new technological revolution, after the Internet. Today, universities have already started training their students to learn how to use the different AIs available on the market.
While new generations learn to familiarize themselves with these tools, researchers are experimenting to create new technologies boosted by AI. This is particularly the case of experts from the American University of Perdue. Indeed, they have just presented their work on a new AI capable drastically improve the performance of thermal and infrared imaging.
See at night as in broad daylight, this is the promise of this AI
The goal ? Allowing our devices to see at night as in broad daylight. Named HADAR for Heat Assisted Detection and Ranging, this technology combines thermal detection, infrared imaging and machine learning. According to the researchers, this AI can detect objects and determine their distance with an accuracy ten times greater than that offered by current night vision technologies.
“Hadar strikingly retrieves the texture of the thermal signal and accurately unravels the temperature, emission, and texture, or TeX, of all objects in a scene. It sees texture and depth in darkness as if it were daytime and also perceives physical attributes beyond visible RGB (red, green and blue) imaging or conventional thermal sensing. It is surprising to find that it is possible to see in total darkness as well as in broad daylight”, summarizes Fanglin Bao, one of the project’s researchers.
Also read: Thanks to this AI, Google will soon allow us to speak naturally to robots
As you can imagine, the possible applications of this technology are endless and could relate to many areas: the defense of course (the project is funded directly by DARPA, the American agency in charge of R&D of technologies for military use), without forgetting transportation such as automobile (autonomous cars in particular), the robotics, agriculture, health or even theexploration of the seabed For example.
For now, researchers cannot yet miniaturize the prototype to integrate it into various products, not to mention its exorbitant cost. Moreover, its execution speed still leaves something to be desired, since it takes a second to create an image. In the case of use in an autonomous car, it would have to produce at least 30 frames per second to be really effective.
Source: University of Perdue