Scientists at Purdue University and Los Alamos National Laboratory have developed what they call “heat-assisted detection and ranging” (HADAR), which combines commercial infrared cameras with machine learning algorithms to more accurately create nighttime vision.
Although objects continuously emit thermal radiation, the particles quickly diffuse into the environment, which is why thermal imaging is usually blurry, an effect called ‘ghosting’. To solve this issue, the researchers trained an AI algorithm specifically to distinguish between objects and their surroundings in images captured by commercial infrared cameras.
HADAR is fundamentally different, it uses invisible infrared radiation to reconstruct a night-time scene with clarity like daytime.
Zubin Jacob, Electrical and Computer Engineering Professor at Purdue University
“Active modalities like sonar, radar and LiDAR [Light Detection and Ranging] send out signals and detect the reflection to infer the presence/absence of any object and its distance. This gives extra information of the scene in addition to the camera vision, especially when the ambient illumination is poor”, Zubin Jacob, a professor of electrical and computer engineering at Purdue and article co-author, told PopSci.
Thus, HADAR can create detailed images by detecting temperature, material composition and thermal radiation patterns in any conditions, be it nighttime or other visual impediments like smoke or fog. “HADAR not only sees texture and depth through the darkness as if it were day but also perceives decluttered physical attributes beyond RGB or thermal vision, paving the way to fully passive and physics-aware machine perception”, reads the abstract of the research paper.
Our work leads to a disruptive technology that can accelerate the Fourth Industrial Revolution with HADAR-based autonomous navigation and human–robot social interactions.
Bao et al., Heat-assisted detection and ranging
The researchers are looking for real-life applications of the HADAR technology. It has the potential to not only assist human drivers when they are driving at night, but it could also be used to power self-driving cars. Other foreseen applications are on autonomous robots or touchless security screenings.
“It is great to know that thermal photons carry vibrant information in the night similar to daytime. Someday we will have machine perception using HADAR which is so accurate that it does not distinguish between night and day”, Jacob said.
The scientists did admit that the road to self-driving nighttime cars is rather long as currently the technology is too expensive for commercial vehicle implementation. Moreover, environmental factors can still interfere with HADAR’s accuracy, which also requires real time calibration.
