Road-Network-Based Fast Geolocalization

In this article, a road-network-based geolocalization method is proposed. We match roads in the onboard images to reference road vector map, and realize successful localization over areas as large whole city. The network matching problem treated point cloud registration under homography transformation solved hypothesize-and-test framework. To tackle problem, global projective-invariant feature proposed, which consists of two intersections augmented with their tangents. addition, we propose necessary conditions for features match. This can reduce candidate features, thus accelerating search great extent. These candidates are first “filtered” model consistency check parameter space then tested similarity metrics identify correct transformation. experiments show that our localize an aerial image area larger than 1000 km 2 within several seconds on single CPU. Our code be found at: https://github.com/FlyAlCode/RCLGeolocalization-2.0.