Google Patents a System That Labels Only the Landmarks You Can Actually See
Google is patenting a way to make map labels smarter by only showing you the names of things that are actually in your line of sight, not every building within a mile radius.
How Google figures out which buildings you can see from where you are
Imagine you're standing on a busy street corner holding up your phone to look at a map. The app floods your screen with labels for dozens of nearby restaurants, hotels, and landmarks, including ones blocked by the building right in front of you. That's the problem Google is working to fix.
This patent describes a system that pre-calculates which structures are visible from thousands of specific spots across a city. It builds a kind of invisible grid over the map, and for each square in that grid, it records exactly which building faces and landmarks you could see from there. When you open the app, it figures out which grid square you're standing in and pulls up only the labels that match what's physically visible from your position.
The result is a map overlay that reflects the real world in front of you. Instead of being buried in labels for things you can't see, you'd get annotations for the actual buildings in your current view.
How visibility cells map 3D models to real-world locations
The system starts with 3D models of buildings and structures in a city, the kind of detailed geometry that Google already collects for products like Maps and Earth.
It then divides the city into a grid of map projection cells (think of tiny geographic tiles, like squares on a chessboard, each representing a real patch of ground). For every cell in that grid, the system runs calculations to determine which parts of which buildings are actually visible from that spot, taking into account walls, trees, and other obstructions. The results are stored as visibility data, essentially a pre-computed lookup table linking each grid square to the building faces you could see while standing there.
When a user's device sends view data, information about the current location and camera orientation, the system matches it to the right grid cell and consults the pre-built visibility table. It then cross-references a point of interest database to find which named places (restaurants, landmarks, transit stops) sit on the visible building faces, and generates labels only for those.
Pre-computing the visibility offline is the key design choice here. Doing that math live on a phone for every step you take would be slow; doing it once on servers and caching the results makes real-time annotation practical.
What this means for Google Maps and street-level navigation
For everyday navigation, this kind of precision would make augmented-reality map overlays far less cluttered. AR navigation, where labels float over a live camera view of the street, already exists in Google Maps in limited form. The main complaint is visual noise: too many labels crowding the screen. A visibility-aware system would automatically thin that out.
It also hints at how Google is thinking about city-scale 3D data as infrastructure. If visibility grids are pre-built for entire cities, the same underlying data could power other features beyond map labels, things like routing recommendations based on what's in sight or context-aware search results tied to what your camera is pointed at.
This is a quiet but genuinely useful engineering problem. AR navigation's biggest UX failure right now is label overload, and Google is tackling it with a systematic, pre-computation approach rather than a quick filter. It's not flashy, but it's the kind of foundational work that makes consumer features actually usable.
Get one Big Tech patent every Sunday
Plain English, intelligent commentary, no hype. Free.
Editorial commentary on a publicly published patent application. Not legal advice.