%0 Conference Proceedings
%T Accurate location of façades of interest in street view panoramic sequences
%D 2015
%A Araujo, André Alvarado,
%A Sampaio, Jonas da Costa,
%A Evangelista, Raphael dos Santos,
%A Mantuan, Altobelli de Brito,
%A Fernandes, Leandro Augusto Frata,
%@affiliation Universidade Federal Fluminense (UFF)
%@affiliation Universidade Federal Fluminense (UFF)
%@affiliation Universidade Federal Fluminense (UFF)
%@affiliation Universidade Federal Fluminense (UFF)
%@affiliation Universidade Federal Fluminense (UFF)
%E Papa, João Paulo,
%E Sander, Pedro Vieira,
%E Marroquim, Ricardo Guerra,
%E Farrell, Ryan,
%B Conference on Graphics, Patterns and Images, 28 (SIBGRAPI)
%C Salvador
%8 Aug. 26-29, 2015
%I IEEE Computer Society
%J Los Alamitos
%S Proceedings
%K street view, geographical coordinates, environment map, feature extraction, feature matching, error propagation.
%X Geo-spatial queries, i.e., queries that combine location data with other kinds of input, have taken huge importance in the last generation of search engines. The success of a geo-spatial search depends on the quality of the positioning information provided, for instance, by GPS-enabled smartphones. Therefore, the quality of the GPS signal and the quality of the built-in GPS may affect the accuracy of the estimated location, and hence the quality of the searching result. This paper proposes an automatic image-based solution for improving the estimation of the geographical coordinates of a building of interest on which a geo-spatial search will be performed. Our approach uses the inaccurate GPS coordinates estimated by smartphones as starting point for automated visual search into a graph of streets enhanced with street view panoramic sequences. During the search, our approach uses a query image of the building of interest to identify which panoramic views include the building's façade. From the geographical location of the panoramic views and from the best matching directions of the given image with the panoramic images, our approach triangulates the location of the target building. In addition, our approach estimates the uncertainty in the computed locations by modeling the error propagation along the triangulation procedure. We evaluate our method on several real images of buildings.
%@language en
%3 PID3770445.pdf