The methodology of building a digital elevation model based on the results of aerial photogrammetric survey from an unmanned aircraft is proposed, which is based on the division of the initial point cloud into equal segments. This allows, having made an assumption of the linear character of change of height of points in a separate segment, to approximate them by separate planes. RMS errors of the models from the survey data were calculated according to the scattering of the points in relation to the approximating surfaces, which made it possible to reveal the dependence of the model construction error relative to the sizes of their constituent segments, as well as to propose a method for filtering the cells containing outliers with respect to the expected model error. The proposed method was tested on the models of three mining objects – limestone quarry, phosphogypsum dump, and peat cut. The experimental results showed a multiple reduction in model error compared to standard DEM models providing the required accuracy for mining documentation.