Fast extraction of buildings with high accuracy from remote sensing images is an important research of remote sensing intelligent application services. To address the problems of imprecise segmentation of building edge in remote sensing images, holes in large-scale target segmentation, and a large amount of network parameters in the DeepLab model, a lightweight DeepLabv3+ model for building extraction from remote sensing images is proposed. In this method, the lightweight network MobileNetv2 is used to replace Xception, the backbone network of DeepLabv3+, so as to reduce the number of parameters and improve the training speed; The hole rate of hole convolution in ASPP is optimized to improve the effect of multi-scale semantic feature extraction. The improved model has been tested on WHU and Massachusetts data sets. The results show that the IOU and F1 score in WHU dataset are 82.37% and 92.89%, respectively, 2.71 percentage points and 2.14 percentage points higher than those of DeepLabv3+, 2.04 percentage points, and 2.32 percentage points higher than those of DeepLabv3+ in Massachusetts data set. The number of training parameters and training time is reduced, and the accuracy of the building extraction is effectively improved, which can meet the requirements of fast extraction of high-precision buildings.
Fan R S, Chen Y, Xu Q H, et al. A high-resolution remote sensing image building extraction method based on deep learning[J]. Acta Geodaetica et Cartographica Sinica, 2019, 48(1):34-41.
Zhang H R, Zhao J H, Zhang X G. High-resolution image building extraction using U-net neural network[J]. Remote Sensing Information, 2020, 35(3):143-150.
Xu Z Y, Shen Z F, Li Y, et al. Building extraction by deep learning method combined with ambiguity and morphological index constraints[J]. Journal of Geo-Information Science, 2021, 23(5):918-927.
Tang Y, Liu Z J, Yang Y, et al. Research on building extraction based on neural network with feature enhancement and ELU activation function[J]. Journal of Geo-Information Science, 2021, 23(4):692-709.
Liu Y F, Zhang Q Z, Wang G H, et al. Building extraction in remote sensing imagery based on deep residual network[J]. Remote Sensing Information, 2020, 35(2):59-64.
Chen K Q, Gao X, Yan M L, et al. Building extraction in pixel level from aerial imagery with a deep encoder-decoder network[J]. Journal of Remote Sensing, 2020, 24(9):1134-1142.
Ye Y X, Tan X, Sun M M, et al. High-resolution remote sensing image classification based on improved DeepLabV3 network[J]. Bulletin of Surveying and Mapping, 2021(4):40-44.
[8]
Lecun Y, Bottou L. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11):2278-2324.
doi: 10.1109/5.726791
[9]
Krizhevsky A, Sutskever I, Hinton G. ImageNet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25(2):1097-1105.
[10]
Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. Computer Science, 2014:1409-1556.
[11]
He K M, Zhang X Y, Ren S Q, et al. Deep residual learning for image recognition[EB/OL].(2015-12-10)[2021-05-12]. https://arxiv.org/abs/1512.03385.
[12]
Iandola F N, Han S, Moskewicz M W, et al. SqueezeNet:AlexNet-level accuracy with 50x fewer parameters and <0.5 MB model size[EB/OL].(2016-11-04)[2021-05-12]. https://arxiv.org/abs/1602.07360.
[13]
Tan M, Chen B, Pang R, et al. MnasNet:Platform-aware neural architecture search for mobile[C]// 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition,IEEE, 2019:2820-2828.
[14]
Howard A G, Zhu M, Chen B, et al. MobileNets:Efficient convolutional neural networks for mobile vision applications[EB/OL].(2017-04-17)[2021-05-12]. https://arxiv.org/abs/1704.04861.
[15]
Chen L C, Papandreou G, Kokkinos I, et al. Semantic image segmentation with deep convolutional nets and fully connected CRFs[J]. Computer Science, 2014(4):357-361.
[16]
Chen L C, Papandreou G, Kokkinos I, et al. DeepLab:Semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4):834-848.
doi: 10.1109/TPAMI.2017.2699184
[17]
Chen L C, Zhu Y, Papandreou G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]// Conference on Computer Vision(ECCV), 2018:801-818.
Wang J Q, Li J S, Zhou H C, et al. Typical element extraction method of remote sensing image based on DeepLabv3+ and CRF[J]. Computer Engineering, 2019, 45(10):260-265,271.
Liu W X, Shu Y Z, Tang X M, et al. Remote sensing image segmentation using dual attention mechanism DeepLabv3+ algorithm[J]. Tropical Geography, 2020, 40(2):303-313.
[20]
Sandler M, Howard A, Zhu M, et al. Mobilenetv2:Inverted residuals and linear bottlenecks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2018:4510-4520.
[21]
Chollet F. Xception:Deep learning with depthwise separable convolutions[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition, 2017:1800-1807.
[22]
He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 37(9):1904-1916.
[23]
Ji S P, Wei S Q, Lu M. Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 57(1):574-586.
doi: 10.1109/TGRS.2018.2858817
[24]
Mnih V. Machine learning for aerial image labeling[D]. Toronto: University of Toronto, 2013.