The Team Conducts 3D Recognition Of Transparent Objects In Under Two Seconds

What once required 15 seconds now takes under 2: Thanks to innovative single-shot technology, the goROBOT3D system—developed by researchers at the Fraunhofer Institute for Applied Optics and Precision Engineering IOF—can now measure 3D objects much more efficiently, even those that are transparent or black. The institute will unveil this technology for the first time at the automatica trade show show in Munich, taking place from June 24 to 27.

What is easy for the human eye to see—such as transparent glass, shiny metal, or deep black surfaces—remains a major challenge for robotic systems. These so-called “uncooperative” surfaces are difficult for standard sensors to detect. To tackle this, researchers at Fraunhofer IOF have developed the goROBOT3D system, which leverages intelligent thermal imaging to identify such challenging materials.

Fraunhofer IOF has been advancing thermal 3D sensor technology for years. Their latest breakthrough significantly improves the system’s efficiency for automated object recognition. GoROBOT3D now measures and analyzes transparent or deep black objects in under 1.5 seconds—down from 15—using a newly developed projection method for thermal 3D sensors.

One-shot Capture Replaces Sequential Recording

The newly developed method applies single-shot technology to thermal 3D measurement. Our approach involves heating the surface of the measurement scene in a structured way. This creates a static thermal point pattern on the object’s surface, captured by two thermal cameras. Dr. Martin Landmann from Fraunhofer IOF explains that spatial cross-correlation lets us create a 3D image from the recorded pair.

Instead of traditional fringe projection, the team uses two DOEs to create an irregular point pattern. These DOEs work by diffracting the incoming laser beam, splitting it into a structured pattern. By precisely combining the DOEs, researchers quickly and effectively projected the point pattern onto transparent objects for the first time.

From Point Cloud to Gripping Command In Milliseconds

Previously, processing transparent materials required capturing and analyzing a large number of image pairs—a slow and labor-intensive task. “Our new method replaces the need for hundreds of thermal image pairs with just one,” explains Dr. Landmann. “This allows us to reconstruct 3D data in just a few milliseconds, reducing measurement and evaluation time by a factor of ten.”

Artificial intelligence processes the 3D data to identify optimal gripping points and angles, then relays this information to a robotic arm equipped with a suction gripper. “We apply a bin-picking approach,” says the Fraunhofer researcher, “meaning we can precisely grasp objects even in disordered settings.”

Seamless Production Without Cycle Interruptions

The significantly shortened recording and evaluation times open up new opportunities for automation in industrial settings, such as manufacturing and product development. Robots can now reliably detect and handle transparent or dark items while maintaining continuous workflow. “As one object is being processed, the next can already be measured,” explains Martin Landmann. “This enables seamless production processes.”

Due to its modular design, the single-shot technology can be easily adapted to a variety of applications. Those interested can experience it firsthand at automatica in Munich, taking place from June 24 to 27, 2025. Visit Fraunhofer IOF at the world’s premier trade fair for smart automation and robotics in Hall 4, Booth 319.

Read the original article on: Tech Xplore

Read more: Hugging Face Says Its New Robotics Model Runs Efficiently On A MacBook