Interfacing photographs with geographic databases is a useful approach not only for architects, historians, geographers but also for a general audience. By assigning a geographical position to a picture it becomes possible to describe its potential content based on the information contained in the geographic database.The aim of the thesis is to propose a method to locate automatically an urban photograph by comparing this photograph with 3D images generated systematically from a virtual model of the buildings. The principle is to put in relation a photograph with the synthetic images that are sharing analog geometric characteristics and therefore to deduce the location of the photograph from the known position of the image. In order to nd synthetic images related to a photograph the method used is a matching between detected lines in the photograph and the images based on Hough transform. This matching is followed by a statistical analysis to propose a probable location with an estimation of accuracy. Despite the obstacles present in the photographs, this approach usingonly lines is simple and potentially effective for positioning photographs in urban space. Several scenarios of use are proposed : first location of an image, validation of the position of an image broadly located and use of photographs to identify change in the urban landscape.
