The invention discloses a mining area automatic semantic segmentation method in
remote sensing image in the technical field of
automatic target detection and depth learning, characterized in that: thespecific steps are as follows: 1, establishing a training sample set: acquisition of
Remote Sensing Images of Mining Areas, and artificial delineating the boundaries of the mining area, forming a Boundary
Grid File, 448*448 being generated by ArcGIS, 512*512
fishing nets of two scales, the
remote sensing image of the mining area being
cut in batches by using the two generated
fishing nets to generate image blocks of different sizes as input data of the depth
learning network, and the boundary files of the mining area grid in the image being
cut through the
fishing nets to generate boundary files corresponding to each mining area image block as
label data of the network. The invention adopts a
hybrid network Den-Res Net can abstract the extracted features while preserving the integrity offeatures, which can be used to solve the
redundancy problem of Dense Net network, and the method has high efficiency, automatic semantic segregation and high accuracy.