MIT's new portable imaging system GelSight can achieve instant 3D imaging

Researchers at MIT have developed a simple, portable imaging system called GelSight that combines a transparent plate, synthetic rubber, an outer paint layer with tiny metal particles, and a clever algorithm to enable 3D imaging that was previously only possible with large and expensive laboratory equipment.

The device provides a way to inspect large products that are too bulky to be examined under a microscope, and could also have future applications in medicine, forensics and biometrics.

GelSight places a synthetic rubber on a transparent plate, one side of which is coated with paint containing small metal particles. When the rubber on the plate is pressed against the surface of an object, the painted side of the plate will deform. A camera mounted on the other side of the plate captures the results, and computer vision algorithms analyze these images.

The new, higher-resolution version of GelSight can reach physical features less than 1 micron in depth and about 2 microns in cross section.

GelSight grew out of a project to build a tactile sensor for robots, after the researchers realized their system could provide a higher resolution than was needed for touch sensing.

The researchers reduced the size of the metal particles on the paint surface and used a different lighting scheme than before, which required redesigning the computer vision algorithms to measure surface features.

Traditionally, large and expensive equipment such as confocal microscopes or white light interferometers are required to produce micron-scale images, and it can take minutes to hours to produce a single 3D image. Such equipment usually has to be mounted on a table that can be isolated from vibrations, which may require a granite tabletop and shock absorbers.

In contrast, researchers Edward Adelson and Micah Kimo Johnson built a sensor prototype about the size of a soda can that can generate 3D images almost instantly.

When multiple cameras measure the deformation of the rubber, the system generates a 3D model of the object, which can be examined from multiple angles on a computer screen.

Adelson and Johnson are in discussions with a major aerospace company and several industrial equipment manufacturers who are interested in using GelSight to verify product integrity.

The technology has also attracted great attention from criminal forensics experts, who believe it can provide a low-cost but efficient identification method for ballistic matching of wooden branches.