TECHNOLOGY Disney Research
Engraved panels cast brilliant image Carving thousands of intricate facets into a transparent sheet recreates any image you want Jacob Aron
curved areas that act as lenses to project these bright patches. Weyrich and his team determine the exact pattern of patches required by looking at the energy distribution of a greyscale image: brighter regions have more energy, while darker ones have less. They then work out the collection of curved patches needed to reproduce this energy distribution. If that sounds tricky, manufacturing the required surface is even harder. Each curved patch has to be painstakingly
SUNLIGHT shining through the window could one day project your favourite picture onto the wall, thanks to a new system that generates complex images by carving intricate patterns into the surface of a transparent sheet. Shine a light through a wine glass and you’ll see refracted slivers of brightness overlaying the glass’s shadow. It is complex versions of these bright patterns, called caustics, that are now being exploited to reproduce “Over 1000 curved areas a photographic image. act as lenses, shaping Tim Weyrich, a computer the light into fuzzy graphics researcher at University elliptical patches” College London, worked with researchers at Disney Research in Zurich, Switzerland, and Princeton carved out by a computercontrolled mill, and producing University to manufacture a single slab can take up to three Plexiglas slabs that generate an days. Weyrich hopes this can array of fuzzy elliptical patches which together form a predefined eventually be speeded up. Another limitation with the image. Each 10-centimetre-square current method is that lighter slab contains over 1000 tiny
Garbage-sorting robot gets its hands dirty A ROBOT that automatically categorises waste from construction and demolition projects could enable valuable raw materials to be recycled instead of ending up in landfill. Industrial robots normally excel at precise tasks in controlled environments, such as assembling cars. More chaotic and hazardous tasks have fallen to humans – for example, sorting through piles of 22 | NewScientist | 2 April 2011
waste in search of precious raw materials that could be recycled. But perhaps not for much longer. Led by Tuomas J. Lukka, a team of researchers at ZenRobotics in Helsinki, Finland, are hoping robots can take over waste recycling. The company’s Recycler robot uses data from a combination of visual sensors, metal detectors, weight measurements and tactile feedback from a robotic arm to pick out likely pieces of refuse and categorise them. Through trial and error its machine learning software has been taught to recognise around a dozen types of material, including different plastics.
–Through a glass brightly–
regions show more detail because they are made up of smaller patches in which the light is more concentrated, compared with the larger patches of more diffuse light making up the dimmer regions. This effect could be avoided by using differently sized carvings in the glass – large ones to focus large amounts of light for highlight areas and smaller ones for the shadows – but that would make designing the surface still more complicated. Weyrich and colleagues have also applied the techniques to reflected light by manufacturing metallic surfaces that generate highlights in the shape of a
desired image. These could be used as a security feature similar to the holograms now imprinted on credit cards, because they would be hard to forge or copy. Weyrich’s system is unlikely to replace simple projectors for producing an image, says Gustavo Patow, a computer graphics researcher at the University of Girona, Spain. But it could find other uses. Car manufacturers could exploit the technique to shape headlight beams into an exact pattern on the road, avoiding the risk of driver glare, Patow suggests. “Being able to control it so finely is amazing.” n
And it can pluck out concrete, metal and wood from a stream of waste as it moves along a conveyor belt. “I’ve never heard of anyone actually trying to do this in such an unstructured environment,” says Edwin Olson, a computer scientist at the University of Michigan in Ann Arbor. For more ambiguous types of waste, such as a piece of plywood with nails driven through it, the robot uses a spectrometer to recognise objects based on the unique patterns of light they reflect. This means the robot can distinguish the type of waste based on its colour and
drop it into the appropriate bin. Since the launch of the test phase in February, the robot has learned to correctly identify half of the construction debris it is fed. That’s far from perfect. But in the US, construction waste accounts for 50 per cent of all landfill material, according to the Construction Materials Recycling Association. Recycling just a fraction of that would mean big savings in resources, as well as landfill fees. Though engineered for construction waste, the robot could one day sort household waste as well, says Lukka. Zena Iovino n