Abstract: The problem of surface subsidence caused by high-intensity underground mining of coal has become increasingly
prominent,and horizontal displacement is a key factor leading to uneven surface deformation and geological disasters. Aiming
at the problems of low efficiency and limited coverage of traditional observation methods,a global observation method of
surface horizontal displacement based on UAV photogrammetry and feature recognition algorithm is proposed. Taking the surface
of 12308 working face of Hongqingliang Coal Mine in Ordos City,Inner Mongolia as an example,the measurement accuracy
and stability of three feature recognition algorithms,FAST,Harris and SIFT,are systematically compared. The results show
that the horizontal displacement observation error of SIFT algorithm outside the mining influence range is 1. 20-2. 08 cm,which
is significantly better than FAST algorithm (2. 06-3. 09 cm) and Harris algorithm (1. 53-3. 82 cm). In the range of mining
influence,the error of SIFT algorithm is controlled in the range of 0. 67-2. 18 cm,which is also better than FAST and Harris
algorithms. Its multi-scale feature extraction ability,scale invariance and strong robustness to illumination change and vegetation
interference can ensure stable recognition of feature points under complex surface topography without loss of feature points.
Compared with the traditional single-point observation method,the proposed method can achieve global coverage observation with centimeter-level accuracy,and has the advantages of high-frequency dynamic update and low cost. By improving the resolution
of UAV images,optimizing the multi-scale feature fusion strategy and constructing a dynamic feature point database,the
adaptability of the algorithm in noisy environment can be further improved. The research results provide a new method for the
observation of surface subsidence and geological disasters in the mining area,and have important application value for ensuring
the safe mining and ecological protection of the mine.

Key words: mining subsidence,photogrammetric survey,feature identification,matching algorithm,digital orthoimage