Abstract: Surface subsidence information in mining areas can be obtained efficiently by low-altitude UAV aerial imagery via modelling overlay. However, vegetation cover and systematic errors in aerial survey often lead to excessive noise in the subsidence model, which limits the application of aerial survey technology in mining subsidence monitoring. To this end, taking the coal mining subsidence area of a mine in northern Shaanxi Province as the experimental region, this study constructs an initial subsidence model using low-altitude UAV aerial imagery data. On the foundation of analyzing the influence of major features such as vegetation, roads and sand on the subsidence modelling accuracy by selecting samples from non-subsidence areas, the vegetation covered area subsidence data which had great influence on the accuracy of subsidence model were removed from the initial subsidence model using the Visible Difference Vegetation Index (VDVI). So a surface subsidence model with vegetation influence removed is then generated, by combining a Gaussian convolution kernel with a digital elevation model interpolation algorithm to fit the subsidence information. According to the prior condition that the amount of subsidence in the non-subsidence area should be zero, the potential systematic errors in the subsidence model are extracted from the statistical error distribution characteristics of the non-subsidence area, and the distance between the statistical sample area and the control point is taken as the weight to analyze the propagation law of systematic errors, and the systematic error correction is then applied to the subsidence model. The comparison of the measured data shows that the accuracy of the subsidence model has been significantly improved after vegetation removal and the subsidence curve of the main section of the subsidence model is more consistent with the measured data after systematic error correction, that can meet the basic requirements of surface subsidence monitoring for large scale and high intensity coal mining in western low vegetation cover areas. The research results show that the UAV image modeling method after removing the noise and correcting the influence of systematic error can meet the basic requirements of large-scale and high-intensity coal mining surface subsidence monitoring in the western low vegetation coverage area, which provides a feasible scheme for the UAV remote sensing technology to be used for efficient monitoring and fine modeling of coal mining subsidence in the western mining area.