-
The new Google Maps pulls in color data from satellite imagery, so you can now see things like Mount Rainier. [credit: Google / Ron Amadeo ]
Don't be too surprised if Google Maps suddenly starts looking different soon. Google announced a rework for the Google Maps normal map tileset which will now pull in color data from satellite imagery. Google published a few before-and-after comparison shots, which we've aligned with the source satellite images in the gallery above.
The company is also tweaking the color palette for a more vibrant map—ocean blues are bluer, forest greens are greener, and the map now even shows snowy mountain peaks in white and barren land in brown. The world is a big place, so of course, this is all happening algorithmically. Google explains:
Google Maps has high-definition satellite imagery for over 98 percent of the world’s population. With a new color-mapping algorithmic technique, we’re able to take this imagery and translate it into an even more comprehensive, vibrant map of an area at global scale.
How exactly does this color-mapping technique work? First, we use computer vision to identify natural features from our satellite imagery, looking specifically at arid, icy, forested, and mountainous regions. We then analyze these features and assign them a range of colors on the HSV color model. For example, a densely covered forest can be classified as dark green, while an area of patchy shrubs could appear as a lighter shade of green.
Previously, the base map was only ever grey (normal land), green (vegetation and parks), or blue (water). The big addition is the coloring of dirt and snow. It's not exactly clear what's going on with the normal ground color. Arizona is now very brown, which makes sense as Google says it now "accurately reflects its desert landscape." But another slide shows Morocco and a chunk of the Sahara Desert, but that's... still gray? It seems like there are still some bugs to work out.
Read 1 remaining paragraphs | Comments
https://ift.tt/2YdY8IS
Comments
Post a Comment