Indoor localization with a signal tree

Indoor localization based on image matching faces the challenges of clustering large amounts of images to build a reference database, costly query when the database is large and indistinctive image features in buildings with unified decoration style. We propose a novel indoor localization algorithm using smartphones where WiFi, orientation and visual signals are fused together to improve the localization performance. The reference database is built as a signal tree with less computational cost as WiFi and orientation signals pre-cluster the reference images. During localization, WiFi and orientation signals not only offer more context information, but also prune impossible reference images, improving the accuracy and efficiency of image matching. In addition, images are described by multiple-level descriptors recording both global and local image information. The proposed method is compared with other methods in terms of localization accuracy, localization efficiency and time cost to build the reference database. Experimental results on four large university buildings show that our algorithm is efficient and accurate for indoor localization.