Google Maps is a practical application but still perfectible. The Mountain View company admits it itself in a blog note and announces work to improve one of the most important features of its service, but sometimes knows failures: the accuracy of GPS. And yes, it has already happened to everyone: you are on Maps and the blue dot is not positioned in the right place, or the halo indicating which direction you are shot is in the wrong direction. The goal is to correct these kinds of problems.
Google Maps will have a more accurate GPS
Google explains that currently, Maps determines your position and your orientation thanks to two main tools: GPS technology and the compass. But "both have physical limitations that make it difficult to solve this problem [of vagueness], especially in urban areas," says the company. That's why this one looked for a new solution, and think to have found it with what it calls the "Global Localization".
It is a process using Visual Positioning Service (VPS), Street View, as well as machine learning and therefore AI. The VPS "first creates a map by taking a series of well-known images and analyzing them for vital visual elements, such as the outline of buildings or bridges, to create a picture. large-scale and rapidly searchable catalog of all these visual elements. "
Obviously, in this way, the accuracy of the location depends strongly on the quality of the imagery. This is where Street View, a source of high-definition photos of environments, comes into play. It is this service that allows Google Maps to offer augmented reality navigation that we talked about on February 11th. As for AI, it plays many roles. Its main is probably to identify the most important elements of the decor, those that are not likely to disappear, to be altered or changed over time. As we know, Street View photos are not always recent and environments can change. All this is still in development, but looks very promising and Google seems convinced that this is the ultimate answer to the inaccuracy of GPS navigation apps.
