Abstract
Extracting buildings and roads from remote sensing images is very important in the area of land cover monitoring, which is of great help to urban planning. Currently, a deep learning method is used by the majority of building and road extraction algorithms. However, for existing semantic segmentation, it has a limitation on the receptive field of high-resolution remote sensing images, which means that it can not show the long-distance scene well during pixel classification, and the image features is compressed during down-sampling, meaning that the detailed information is lost. In order to address these issues, Hybrid Multi-resolution and Transformer semantic extraction Network (HMRT) is proposed in this paper, by which a global receptive field for each pixel can be provided, a small receptive field of convolutional neural networks (CNN) can be overcome, and the ability of scene understanding can be enhanced well. Firstly, we blend the features by branches of different resolutions to keep the high-resolution and multi-resolution during down-sampling and fully retain feature information. Secondly, we introduce the Transformer sequence feature extraction network and use encoding and decoding to realize that each pixel has the global receptive field. The recall, F1, OA and MIoU of HMPR obtain 85.32%, 84.88%, 85.99% and 74.19%, respectively, in the main experiment and reach 91.29%, 90.41%, 91.32% and 84.00%, respectively, in the generalization experiment, which prove that the method proposed is better than existing methods.
Highlights
Land resources have the following qualities as carriers of human existence and development: nonrenewable resources, fixed location, and imbalanced distribution [1]
It can be seen that the current semantic segmentation networks have limitations in the receptive field, so this work combines the Transformer method with global receptive fields [24,25,26,27,28] to deeply mine the semantic information of the feature maps
This paper proposes the Hybrid Multi-resolution and Transformer semantic extraction Network (HMRT) to extract buildings and roads from high resolution remote sensing images
Summary
Land resources have the following qualities as carriers of human existence and development: nonrenewable resources, fixed location, and imbalanced distribution [1]. Aiming at the problem of loss of detailed information during the down-sampling process, Sun et al [13] offered a new feature fusion strategy based on a full convolutional network with ultra-high resolution image segmentation, which maximized the fusion of deep-level semantic features and shallow-level detail information Combining with this model, the effective digital surface model was proposed, and the information of high-resolution remote sensing images was extracted, which improved the accurate segmentation of the full convolutional network. A larger receptive field can be obtained by the use of hollow convolution and feature pyramids, the receptive field is still local, understanding of long-distance scenes cannot be achieved, and pixels cannot be classified precisely To solve these problems, a Hybrid Multi-resolution and Transformer semantic extraction Network (HMRT) is proposed in this study. The small receptive field of the convolutional neural network can be overcome and the understanding of a long-distance scene can be improved. (3) Feature Channels Maximum Element is proposed to strengthen the class location information, which can effectively improve the accuracy of segmentation
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
