Researchers have developed a GPS system, which when constructing a route analyzes satellite images to more accurately determine the type of roads and the number of traffic strips.
Usually, GPS cards create large companies such as Yandex and Google, sending special cars with cameras to fix the infrastructure details of various areas. However, this method of collecting information is expensive and takes a lot of time, so some parts of the world are simply ignored, and the data obtained are basic. In addition, the system does not always correctly define the number of traffic strips and where they lead, so sometimes on the way the route can change dramatically.
One of the options for solving these problems is the use of machine learning algorithms for road recognition on satellite clocks. Although this method is cheaper, gives more information and uses regularly updating images, but roads are often hidden by trees and buildings, which complicates the work of the AI.
Recently, researchers from the Massachusetts Institute of Technology and Qataria jointly developed a combination of neural architectures for automatic forecasting of road types and the number of obstacle movement strips. During testing, the system correctly determined the types of roads in 93% of cases, and the number of traffic strips with an accuracy of 77%.
In the future, the team plans to teach the system to recognize parking spaces, cycles and provide information on the current state of roads.
The efficiency of such systems is still important for unmanned vehicles, which moves with test tracks on
To develop the channel, your support is important to us, subscribe to the channel and put like.