Abstract: Aiming at the problems of low efficiency and strong subjectivity in the traditional manual interpretation method
for identifying abandoned mining subsidence areas,a new method for identifying abandoned mining subsidence areas based on
Unmanned Aerial Vehicle (UAV) photogrammetry and deep learning algorithms is proposed. Firstly,the DJI Phantom 4 RTK
UAV is used to conduct oblique photography in a mine subsidence area to obtain high-resolution image data and 3D point cloud
data. Secondly,the Context Mapper 3D reconstruction software is used to process the obtained data to generate the DOM image
(with a resolution of 2 cm) and DSM data (with a resolution of 5 cm) of the survey area. Finally,a subsidence area identification
model based on deep learning is constructed by combining morphological and spectral features to achieve efficient and accurate
automatic identification of subsidence areas. The experimental results show that the overall accuracy of the proposed
method in identifying abandoned mining subsidence areas reaches 94. 3%,with a Kappa coefficient of 0. 91,which is significantly
better than the traditional manual interpretation method. In addition,compared with the traditional method,the working
efficiency of the proposed method is higher,significantly reducing the influence of subjective factors on the identification results.
Among the 50 validation samples,the proposed method successfully identified 47 subsidence areas,with only 2 missed
and 1 misjudged,demonstrating the high reliability of the method. This provides a new technical path for the rapid identification
and dynamic monitoring of geological disasters and has good application value.