国家自然科学基金(61725105)
[1] Gamba P, Houshmand B, Saccani M. Detection and extraction of buildings from interferometric SAR data. IEEE Trans Geosci Remote Sens, 2000, 38: 611-617 CrossRef ADS Google Scholar
[2] Bolter R. Reconstruction of man-made objects from high resolution SAR images. In: Proceedings of 2000 IEEE Aerospace Conference, 2000. 287--292. Google Scholar
[3] Zhao L J, Zhou X G, Kuang G Y. Building detection from urban SAR image using building characteristics and contextual information. EURASIP J Adv Signal Process, 2013, 2013(1): 56 CrossRef ADS Google Scholar
[4] Fu X Y, You H J, Fu K. 3-D reconstruction of building from single high-resolution SAR image. J Infrared Millim W, 2012, 31: 569--576 [傅兴玉, 尤红建, 付琨. 单幅高分辨率SAR图像建筑三维模型重构. 红外与毫米波学报, 2012, 31: 569--576. Google Scholar
[5] Xu F, Jin Y Q. Automatic Reconstruction of Building Objects From Multiaspect Meter-Resolution SAR Images. IEEE Trans Geosci Remote Sens, 2007, 45: 2336-2353 CrossRef ADS Google Scholar
[6] Pontes J K, Kong C, Sridharan S, et al. Image2mesh: a learning framework for single image 3D reconstruction. In: Proceedings of Asian Conference on Computer Vision. Berlin: Springer, 2018. 365--381. Google Scholar
[7] Kurenkov A, Ji J, Garg A, et al. Deformnet: Free-form deformation network for 3D shape reconstruction from a single image. In: Proceedings of IEEE Winter Conference on Applications of Computer Vision, 2018. 858--866. Google Scholar
[8] Kniaz V V, Remondino F, Knyaz V A. Generative Adversarial Networks for Single Photo 3d Reconstruction. Int Arch Photogramm Remote Sens Spatial Inf Sci, 2019, XLII-2/W9: 403-408 CrossRef ADS Google Scholar
[9] Mandikal P, Murthy N, Agarwal M, et al. 3DLMNet: latent embedding matching for accurate and diverse 3D point cloud reconstruction from a single image. 2018,. arXiv Google Scholar
[10] Fan H Q, Su H, Guibas L J. A point set generation network for 3d object reconstruction from a single image. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017. 605--613. Google Scholar
[11] LeCun Y, Boser B, Denker J S. Backpropagation Applied to Handwritten Zip Code Recognition. Neural Computation, 1989, 1: 541-551 CrossRef Google Scholar
[12] Lecun Y, Bottou L, Bengio Y. Gradient-based learning applied to document recognition. Proc IEEE, 1998, 86: 2278-2324 CrossRef Google Scholar
[13] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks. In: Advances in Neural Information Processing Systems. 2012. 1097--1105. Google Scholar
[14] Zeiler M D, Fergus R. Visualizing and understanding convolutional networks. In: Proceedings of European Conference on Computer Vision. Berlin: Springer, 2014. 818--833. Google Scholar
[15] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. 2014,. arXiv Google Scholar
[16] He K M, Zhang X Y, Ren S Q, et al. Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016. 770--778. Google Scholar
[17] Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015. 3431--3440. Google Scholar
[18] Duan Y P, Tao X M, Han C Y, et al. Multi-scale convolutional neural network for SAR image semantic segmentation. In: Proceedings of IEEE Global Communications Conference, 2018. 1--6. Google Scholar
[19] Xu F, Wang H P, Jin Y Q. Deep learning as applied in SAR target recognition and terrain classification. J Radars, 2017, 6(2): 136-148 DOI: 10.12000/JR16130. Google Scholar
[20] Badrinarayanan V, Kendall A, Cipolla R. Segnet: a deep convolutional encoder-decoder architecture for image segmentation. 2015,. arXiv Google Scholar
[21] Niu S R, Qiu X L, Peng L X, et al. Parameter prediction method of SARtarget simulation based on convolutional neural networks. In: Proceedings of European Conference on Synthetic Aperture Radar, 2018. 1--5. Google Scholar
[22] Xiang Y, Mottaghi R, Savarese S. Beyond PASCAL: a benchmark for 3D object detection in the wild. In: Proceedings of the 2014 IEEE Winter Conference on Applications of Computer Vision. New York: IEEE, 2014. 75--82. Google Scholar
[23] Chang A X, Funkhouser T, Guibas L, et al. Shapenet: an information-rich 3D model repository. 2015,. arXiv Google Scholar
[24] Wu N Z, Song S, Khosla A, et al. 3D ShapeNets: a deep representation for volumetric shape modeling. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015. 1912--1920. Google Scholar
[25] Lim J J, Pirsiavash H, Torralba A. Parsing Ikea objects: fine pose estimation. In: Proceedings of the IEEE International Conference on Computer Vision, 2013. Google Scholar
[26] Choy C B, Xu D, Gwak J Y, et al. 3D-R2N2: a unified approach for single and multi-view 3D object reconstruction. In: Proceedings of European Conference on Computer Vision. Berlin: Springer, 2016. 628--644. Google Scholar
[27] Wang N Y, Zhang Y D, Li Z W, et al. Pixel2Mesh: generating 3D mesh models from single RGB images. In: Proceedings of the European Conference on Computer Vision. Berlin: Springer, 2018. 55--71. Google Scholar
Figure 1
Overall framework of 3D building reconstruction for SAR images based on deep neural network
Figure 2
Architecture of complex valued deep semantic segmentation network for target detection in SAR images
Figure 3
Complex valued convolution layer based on coupling calculation method
Figure 4
Pooling and up-pooling layer based on complex valued data learning
Figure 6
Two groups of SAR image simulation results of typical buildings in different azimuth angles
Figure 7
Point clouds and simulation images generated based on models
Figure 8
Structure of SAR point generation network (SARPGN)
Figure 9
Results of SARPGN on its initial test set
Figure 10
SAR images of (a) Suzhou and (b) Hong Kong
Figure 11
Results of building facade detection in SAR images based on CV-SegNet. (a) SAR images; (b) prediction tags; (c) detection results of fusion of SAR images and prediction tags
Figure 12
Comparison of results of building facade detectionin SAR images. (a) Original amplitude SAR images; protectłinebreak (b) extraction resultsof RV-SegNet; (c) extraction results of CV-SegNet.
Figure 13
Comparisons of semantic segmentation network performance of building facade detection in SAR images. protectłinebreak (a) Overall accuracy; (b) recall; (c) F1 value. The left pictures show the performance of RV-SegNet, and the right pictures show the performance of CV-SegNet. The orange curves in the figures are the results on training set, and the blue curves are the results on test set.
Figure 14
3D models of typical buildings in Hong Kong area. (a) Harbourfront landmark; (b) International Finance Centre; (c) Laguna City; (d) Hyatt Regency; (e) Empire Centre; (f) Hopewell Centre.
Item | RV-SegNet | CV-SegNet |
Overall accuracy | 0.996 | 0.998 |
Recall | 0.900 | 0.920 |
F1 value | 0.920 | 0.940 |