The paper formulates the problem of forming a list of recovery measures to restore elements of the region’s infrastructure from the consequences of natural disasters by automating the identification of problem areas and places that require repair. It is proposed to process information from unmanned aerial vehicles or high-resolution satellite images, using specially trained neural networks, to check the transport infrastructure and the integrity of power lines. Checking the integrity of the transport infrastructure is necessary to ensure that the repair crew can approach the place of rupture or breakdown. If there is no way to get to the repair site, the repair team should be reassigned to another location to keep downtime to a minimum. A neural network has been built and trained, which allows to determine the places of the rubble, fix their coordinates and plot on the map, as well as send the operator photographs of the areas that have raised doubts to correct the information. The neural network allows to determine the location of breaks in power lines and the integrity of the towers. A strategy for compiling a list of repairs is described, which takes into account the places of necessary repairs, access to them, repair time, travel time, time to eliminate congestion and the number of teams available. The results of computational experiments are analyzed.