On-device Scalable Image-based Localization via Prioritized Cascade Search and Fast One-Many RANSAC
It enables GPS-agnostic, on-device localization for mobile users in urban areas, though it is incremental with new optimizations for existing methods.
The paper tackles large-scale urban localization using images on mobile devices without GPS or network, achieving state-of-the-art accuracy on benchmark datasets and demonstrating scalability on a Google Street View dataset.
We present the design of an entire on-device system for large-scale urban localization using images. The proposed design integrates compact image retrieval and 2D-3D correspondence search to estimate the location in extensive city regions. Our design is GPS agnostic and does not require network connection. In order to overcome the resource constraints of mobile devices, we propose a system design that leverages the scalability advantage of image retrieval and accuracy of 3D model-based localization. Furthermore, we propose a new hashing-based cascade search for fast computation of 2D-3D correspondences. In addition, we propose a new one-many RANSAC for accurate pose estimation. The new one-many RANSAC addresses the challenge of repetitive building structures (e.g. windows, balconies) in urban localization. Extensive experiments demonstrate that our 2D-3D correspondence search achieves state-of-the-art localization accuracy on multiple benchmark datasets. Furthermore, our experiments on a large Google Street View (GSV) image dataset show the potential of large-scale localization entirely on a typical mobile device.