Engineers at the University of California, Berkeley, have developed a method to quickly scan and model urban landscapes. From New Scientist


IMAGINE if the first soldiers to enter an enemy city could map it street by street, recording every window and doorway of the urban battlefield in an accurate 3D model that could instantly be relayed to their comrades at base.

Engineers at the University of California, Berkeley, have found a way to do just that. What's more, their technique can also make maps for use by emergency services, urban planners and even tourists looking for the nearest Starbucks.

The concept is similar to building a virtual reality model, but the process is very different. To produce a VR model, a programmer manually combines distance measurements and 2D pictures to make a 3D model. The new technique, dubbed "virtualised reality" by creator Avideh Zakhor, is automated and much faster. "Right now, a detailed urban model can take many months to create," says Bruce Deal, vice-president of the Virginia engineering firm SET Associates, which is helping to adapt the technology for the US military. "With the new model, we're talking about an hour or so." Virtualised reality scans the urban landscape using lasers and digital cameras mounted on a truck or plane. A laser measures distances to objects such as lamp posts and building facades, while the digital camera takes 2D photos. Another laser calculates the movement of the truck and checks its position against data collected from the aerial laser aboard the plane.

These measurements and pictures are fed into a computer that combines them to create a photo-realistic virtual 3D model of the area. Zakhor and her team recently created a working model of downtown Berkeley (see graphic) in just 4½ hours - 26 minutes of driving plus 4 hours of data processing.
“Emergency workers could use the models to figure out the best way to respond to natural disasters or terrorist attacks”

The first user will probably be the US army, which funded much of the research. "Speed makes the system very useful to urban war fighters," says Deal. Each patrol can record new information about its surroundings, updating the model recorded by the previous patrol. Soldiers can keep up with changes to the cityscape, such as new barricades or destroyed buildings. "It's a vast improvement over current military capability," he says. Zakhor has started a spin-off company funded by the Pentagon's Defense Advanced Research Projects Agency to develop an even faster version that creates models in real time.

"The applications are endless," says Zakhor. Car-hire companies or cellphone providers could use similar technology to transmit up-to-date 3D maps to their customers to help them navigate through strange cities. Emergency workers could use the models to figure out the best way to respond to natural disasters or terrorist attacks. Urban planners could even look at a series of models collected over time to see how the layout of their city has evolved.

The process of creating models could be speeded up even further by developments in unmanned aerial vehicles. The US navy is developing cheap (around $2000) robotic aircraft that can operate in "swarms" to perform reconnaissance of a wide area at speed. The aircraft use cooperative software that allows the swarm to cope with some of its members being shot down.

Zakhor's team is also working on a 4D modelling system, which would use lasers and digital cameras to record anything from a boxing match to a ballet performance as a moving 3D model that the viewer could manipulate in virtual space. "You become the director," says Zakhor.