Abstract: Affected by mining subsidence or surface resources stripping，the terrain of mining area is generally complex.In the traditional photogrammetry，a fixed aerial height is designed according to the requirements of average terrain and precision in the mining area，and UAV （Unmanned aerial vehicle） image collection is carried out according to this height. Due to the influence of topographic fluctuation，the photographic scale，image overlap degree and resolution difference in the measurement area are large，resulting in inconsistent measurement accuracy，even can not constitute stereoscopic visual conditions，air triangulation encryption failure.In order to overcome the shortcomings of traditional methods，on the basis of analyzing the influence of topographic relief on measuring accuracy，combing with the terrain following technique，the light and small machine vision module is adopted，the relative altitude of the UAV in real time can be estimated through developing the fast optical flow estimation algorithm based on topology constraints，besides that，the UAV flying height can be adjusted based on the changes of terrain，so as to realize the photogrammetry method of complex terrain in mining areas with real-time terrain tracking.In this way， the uniform photographic scale，image overlap and resolution can be guaranteed under the complex terrain conditions of the mining area，and the uniform photogrammetry can meet the requirements of expected accuracy.The study results show that the accuracy of 1∶500 large scale topographic map can be achieved in the complex terrain of mining area，which proves the effectiveness of the method.

Key words: Mining subsidence, UAV, Photogrammetry, Terrain following, Optical flow method, Complex terrain