A Remote Sensing Land Cover Classification Algorithm Based on Attention Mechanism

摘要

The Deeplabv3+ network for semantic segmentation of remote sensing images has drawbackslike inaccurate edge segmentation and intra-class inconsistency in large-scale segmentation.The attention mechanism is a good solution to this problem. This paper integratesthe dual attention mechanism module including spatial attention mechanism and channelattention mechanism (ISNet) into the network. Furthermore, the ISNet is added to both theencoder part and decoder part in the network. The newly proposed network is calledDISNet. It can enhance target features and suppress background features, having the potentialto improve semantic segmentation accuracy. The experimental results indicate that theMean Intersection over Union (mIoU) of the Deepglobe dataset, GID dataset, and ISPRS testproject (Vaihingen) dataset is respectively 64.22, 62.83, and 78.59%. Those results are 5.68,4.80, and 6.20% higher than that of Deeplabv3+. The proposed network can effectivelyimprove the accuracy of remote sensing land cover classification.