Intersection Detection

A computer-vision or sensor-fusion technique used in indoor navigation systems for blind travellers to identify where two or more walkable corridors meet, so the navigation software can update the user's position on a map and issue a turn instruction at the right moment. Intersection detection typically works either from an image of the surroundings (detecting the shape of the corridor opening with a model such as YOLOv7) or from a 2D occupancy grid built by a depth sensor or LiDAR. The recognised shape — a combination of Left, Right, Front, and Back walkable paths — is matched against the next expected intersection in a planned route. Reliable intersection detection is the foundation for map-less or infrastructure-light indoor navigation for blind users, because it lets a phone localise a user without BLE beacons or prebuilt fingerprints.