Forget X -rays – this new AI radar can identify objects hidden in boxes, loads or under junk
- Advertisement -
Related Posts
- Advertisement -
- MMNorm reconstructs complex hidden forms with the help of Wi-Fi frequencies without touching the object
- Robots can now see in messy drawers using reflected signals from surrounding antennas
- MIT’s technology defeated the current radar protection by 18% over more than 60 objects tested
In environments where visibility is impeded, such as inside cabinets, behind walls or objects, among other things, artificial intelligence could soon have a new way to get ahead.
Researchers from MIT have developed a technique called MM standard, which uses millimeter-wave signals, the same frequency range as Wi-Fi, to reconstruct hidden 3D objects with surprising accuracy.
“We have been interested in this problem for a while, but we have hit a wall because methods in the past, while they were mathematically elegant, did not make us where to go,” said Fadel Adib, senior author of the study and director of the Signal Kinetics Group at MIT.
Radar -overcome restrictions
Previous techniques rely on back projection, which produces and fails images with low resolution when it is applied to small, closed objects such as tools or utensils.
The researchers discovered that the error is in the supervision of a physical characteristic that is known as a specularity – how millimeter golf reflections behave like mirror images.
Instead of easy to measure where signals bounce back, MM standard estimates the direction of the surface, which researchers call the surface normal.
“Trusting on specularity is our idea to try not only to estimate the location of a reflection in the area, but also on the direction of the surface at that time,” Laura Dodds explained, main author on the paper.
By combining many such estimates from different antenna positions, the system reconstructs the 3D curvature of an object, distinguishes between forms that are as nuanced as a mug or the difference between a knife and a spoon in a box.
Each antenna collects reflections with different strength, depending on the orientation of the hidden object.
“Some antennas may have a very strong mood, some can have a very weak voice, and we can combine all voices together to produce one surface normal that is agreed by all antenna belocations,” Dodds added.
This new approach achieved a reconstruction -nominant of 96% over more than 60 objects, which performed better than existing methods that only reached 78%.
The system performed well on objects made of wood, plastic, glass and rubber, although it is still struggling with dense metals or thick barriers.
As researchers work to improve resolution and material sensitivity, the potential use cases grow.
In security scanning or military contexts, MM standard can reconstruct the form of hidden items without opening bags or boxes.
This possibility could be essential for AI-driven robots in warehouse automation, search and salvation, or even assisted living environments.
By Techxplore
Maybe you like it too
- Advertisement -
