According to a report by the Physicist Network, scientists at the University of Cambridge in the UK have recently developed two new systems that allow driverless cars to learn to "see the road." A system called SegNet can determine a variety of objects on the road in real time through a smartphone or a normal camera; another system can identify the user's position and direction in an area where the GPS system cannot provide service.
Developing a driverless car must "teach" it in three ways: Who am I, what is around me, what should I do next. SegNet mainly solves the second problem; another independent system judges its position and orientation by recognizing the image.
The SegNet system takes pictures of road conditions in real time and classifies objects on the road into 12 categories – such as roads, road signs, sidewalks, buildings and cyclists – which function as sensors worth tens of thousands of pounds. For current driverless cars, radar and base sensors are expensive, and in fact they are more expensive than the car itself. Compared to these expensive sensors that use radar to identify objects, SegNet, which uses deep learning techniques, learns through case studies. Researchers used 5,000 images that were accurately labeled to train SegNet to make it available.
Another system requires only a single monochrome photo to locate in a busy city. It is much more accurate than GPS systems and can work in areas where GPS is not available, such as indoors and tunnels, and cities without reliable GPS signals.
Although these two systems are currently not directly available for controlling driverless cars, they can “see†and accurately identify their location and identify the objects they “see†– the key to developing a driverless car problem.
One of the team members, Dr. Alex Kendall, a Ph.D. student at the School of Engineering, said: "The cool thing about research is to use deep learning technology for the first time to let the car judge its position and the surrounding environment." Professor Berto Chappola said: "It still takes time for people to rely entirely on a driverless car, but as the technology is more effective and more accurate, we are closer to the widespread use of driverless cars. â€
Editor-in-chief
When it comes to driverless technology, everyone is most worried about safety. I don’t know if this is the problem that driverlessness has to solve. The computer system that is not fatigued compensates for the driver’s mistakes. The car manufacturer is concentrating on designing An unmanned system that ensures the safety of the car. Today, as mobile phones become the terminal of all things, scientists are opening their minds and discovering the next field in which mobile phones enter. This is not the case, scientists at the University of Cambridge in the United Kingdom have begun to use it and advanced deep learning methods to "church" cars "recognize the road". With these "eyes", in the future without the GPS signal, the driverless technology will also bring We arrived at the destination.
Concerned about surprises
Children Anti Dust Disposable Face Mask
Kids Anti Dust Disposable Mask,Protective Disposable Face Mask,Children Anti Dust Disposable Face Mask,Anti Dust Activated Disposable Face Mask
Guangzhou TaoC E-Commerce Co., Ltd , https://www.aiberrycn.com