The invention discloses a
remote sensing image semantic segmentation method based on
deep learning, and belongs to the technical field of
machine vision. Aiming at the problems of difficulty in obtaining features of small objects and insufficient segmentation precision of a semantic segmentation method of a mainstream deep
convolutional neural network, the method comprises the following steps: improving a Deeplabv3
algorithm, improving a single up-sampling layer, and performing multi-layer up-sampling by utilizing residual errors obtained in a
backbone network to ensure
semantic integrity of an image in resolution; and meanwhile, modifying the
expansion rate of four expansion
convolution layers in the ASPP layer, so that the network has a better effect on small object segmentation. The result shows that the mIou and pixel accuracy of the improved Deeplabv3 semantic segmentation
algorithm on a self-made
data set reaches 94.92% and 98.01% respectively, which are improved by 3.77% and 2.40% respectively compared with the original
algorithm, so that the improved Deeplabv3 semantic segmentation algorithm not only has higher accuracy, but also has better robustness for segmentation of various terrains; the method is suitable for a complex urban
remote sensing image environment, and can be well applied to the fields of
urban planning, agricultural planning, military war and the like.