An Image-based Visual Localization Approach to Urban Space

For positioning problems of the complex environment of urban space, this paper proposed an image-based visual localization approach to urban space by smartphone, realizing of high-precision spatial positioning of the city's complex environment. In the method, firstly, the sequence of the buildings in the prepositioned city space is collected and the SIFT is used for rough matching. The 3D point cloud model of the building is obtained through the optimization of the rough matching point and precisely solving the fundamental matrix and the projection matrix. By the feature points, pixel coordinates and object point coordinates of the object side feature library. Then, a image captured in the city space is used to extract the point features of the location and match with the object feature library to obtain the corresponding object point coordinates, and to accurately calculate the external orientation elements of the city location so as to obtain the location 6 A degree of freedom. Finally, the location of the instant camera position captured by the display terminal is displayed to the positioning user. The experimental results presented in this paper have achieved good spatial positioning of the city and can provide visual positioning accuracy better than 0.1 m.