Two smartphone-based systems that track users outside the visibility of GPS and recognize various objects found on the road scene in real time have been developed. Although these new technology systems have the same function as those high-end sensors currently used, they only cost a fraction of their prices.
Behind these two different but complementary systems are scientists from the University of Cambridge. Demonstration of one of the systems called SegNet can be viewed online for free.
SegNet can capture images on road scene and categorize these on 12 various types whether it is the road, a building or a pedestrian passing by. Researchers claim that its initial tests reveal that it is more efficient than the costly laser and sensors sold and made available in the market.
Although it cannot handle driverless cars yet, these could be a major leap on the automobile technology. Also, it can serve as a warning device just like the anti-collision technologies installed on private cars.
"Vision is our most powerful sense and driverless cars will also need to see," lead researcher Prof. Roberto Cipolla said. "But teaching a machine to see is far more difficult than it sounds."
On the other hand, the second system can find the user's current location even outside the reach of GPS through the use of single color image. Claimed to be better than the satellite signal, the latest system can track even far-flung or indoor areas and can precisely determine whether they are looking at the west or east side. "What's cool about our group is that we've developed technology that uses deep learning to determine where you are and what's around you - this is the first time this has been done using deep learning," Alex Kendall, PhD student from the Department of Engineering, said.
The device was tested during the King's Parade in Cambridge. The system successfully identified the location and orientation from a distance and degree.
As to the systems installation on driverless cars, Prof. Cipolla said that it will take time before people can totally rely on driverless cars, but the more accurate and precise the technologies are being installed, the closer people can easily adopt to the change.
