FEATURES OF AERIAL TRIANGULATION BY USING DIFFERENT-TIME IMAGES OF URBAN AREAS

Abstract

It is well-known that in order to obtain up-to-date spatial information of rapidly developing cities, aerial photography is carried out almost annually. This work is devoted to the joint aerial triangulation of aerial urban area photographs obtained at different times. The main problem of the joint aerial triangulation of different-time images is the process of tie points detection. In this paper it is proposed to search for tie points exclusively on the roofs of building since they are least susceptible to change over time. In order to do this, the roofs of buildings are highlighted on the original aerial photographs via the Unet neural network and then tie points are detected within these areas. The technology made it possible to improve the quality of aerial triangulation: remove photogrammetric gaps in the given block, increase the number of tie points, reduce the processing time by 25% without increasing computing requirements. This approach to the search for tie points made it possible to increase the efficiency of aerial triangulation not only when processing archival and current images together, but also when processing only the results of actual aerial photography.