Evaluating the generalization ability of convolutional neural networks for built-up area extraction in different cities of China

Abstract

The difficulty of build-up area extraction is due to complexity of remote sensing data in terms of heterogeneous appearance with large intra-class variations and lower inter-class variations. In order to extract the built-up area from Landsat 8-OLI images provided by Google earth engine (GEE), we propose a convolutional neural networks (CNN) utilizing spatial and spectral information synchronously, which is built in Google drive using Colaboratory-Keras. To train a CNN model with good generalization ability, we choose Beijing, Lanzhou, Chongqing, Suzhou and Guangzhou of China as the training sites, which are very different in term of natural environments. The ArcGIS-Model Builder is employed to automatically select 99 332 samples from the 38-m global built-up production of the European Space Agency (ESA) in 2014. The validate accuracy of the five experimental sites is higher than 90%. We compare the results with other existing building data products. The classification results of CNN can be very good for the details of the built-up areas, and greatly reduce the classification error and leakage error. We applied the well-trained CNN model to extract built-up areas of Chengdu, Xi’an, Zhengzhou, Harbin, Hefei, Wuhan, Kunming and Fuzhou, for the sake of evaluating the generalization ability of the CNN. The fine classification results of the eight sites indicate that the generalization ability of the well-trained CNN is pretty good. However, the extraction results of Xi’an, Zhengzhou and Hefei are poor. As for the training data, only Lanzhou is located in the northwest region, so the trained CNN has poor image classification ability in the northwest region of China.

References

[1] ChenXH, CaoX, LiaoAP, ChenLJ, PengS, LuM, ChenJ, ZhangWW, ZhangHW, HanG, WuH, LiR. Science China Earth Sciences, 2016, 59: 2295

[2] ZhaY, GaoJ, NiS. International Journal of Remote Sensing, 2003, 24: 583

[3] XuH. International Journal of Remote Sensing, 2008, 29: 4269

[4] PesaresiM, GerhardingerA, KayitakireF. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2009, 1: 180

[5] ChaudhuriD, KushwahaNK, SamalRA, AgarwalC. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2016, 9: 1767

[6] JinX, DavisCH. EURASIP Journal on Advances in Signal Processing, 2005, 2196

[7] PesaresiM, GuoH, BlaesX, EhrlichD, FerriS. IEEE Journal of Selected Topics in Applied Earth Observations & Remote Sensing, 2013, 6: 2102

[8] RanG, StuhlmacherMF, TellmanB, ClintonN, HansonG, GeorgescuM, WangC, S-CandelaF, KhandelwalA K, ChengW H, BallingR. Remote Sensing of Environment, 2018, 205: 253

[9] YangJ, MengQ, HuangQ, SunZ H. A New Method of Building Extraction from High Resolution Remote Sensing Images based on NSCT and PCNN, International Conference on Agro-geoinformatics, 2016, 1

[10] KrizhevskyA, SutskeverI, HintonGE. ImageNet Classification with Deep Convolutional Neural Networks, International Conference on Neural Information Processing Systems, 2012, 60: 1097

[11] SimonyanK, ZissermanA. Very Deep Convolutional Networks for Large-Scale Image Recognition, Computer Science, 2014,

[12] SzegedyC, LiuW, JiaY Q, SermanetP, ReedS, AnguelovD, ErhanD, VanhouckeV, RabinovichA. Going Deeper with Convolutions, 2015, Boston, MA, USA, IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

[13] HeK, ZhangX, RenS, SunJ. Deep Residual Learning for Image Recognition, 2016, 770

[14] CastelluccioM, PoggiG, SansoneC, VerdolivaL. Acta Ecologica Sinica, 2015, 28: 627

[15] VakalopoulouM, KarantzalosK, KomodakisN, ParagiosN. Geoscience & Remote Sensing Symposium, 2015, 50: 1873

[16] HuangZ M, ChengG L, WangH Z, LiH C, ShiL M, PanC H. Building Extraction from Multi-source Remote Sensing Images via Deep Deconvolution Neural Networks, 2016, 1835

[17] MakantasisK, KarantzalosK, DoulamisA, LouposK. Deep Learning-Based Man-Made Object Detection from Hyperspectral Data, 2015,

[18] WanW, MabuS, ShimadaK, HirasawaK, HuJ L. Applied Soft Computing, 2009, 9: 404

[19] Pan X, Luo P, Shi J and Tang X, Two at Once: Enhancing Learning and Generalization Capacities via IBN-Net, Computer Vision, Lecture Notes in Computer Science, vol 11208, Springer, Cham, 2018.

[20] GorelickN, HancherM, DixonM, IlyushchenkoS, ThauD, MooreR. Remote Sensing of Environment, 2017, 202

[21] MartinoP, DanieleE, StefanoF, AnetaF, FreireC, StamatiaH, PierreS, VasileiosS. Operating Procedure for the Production of the Global Human Settlement Layer from Landsat Data of the Epochs 1975, 2014,

[22] LiuX P, HuH, AiB, LiX, ShiQ. Remote Sensing, 2015, 7: 17168

[23] YangN, TangH, SunH Q, YangX. IEEE Geoscience and Remote Sensing Letters, 2018, 5: 257

[24] ZhangT, TangH. Remote Sensing, 2019, 11: 2

[25] ZhangT, TangH. Built-Up Area Extraction from Landsat 8 Images Using Convolutional Neural Networks with Massive Automatically Selected Samples, 2018,

[26] LiuX P, HuG H, ChenY M, LiX, XuX C, LiS Y, PeiF S, WangS J. Remote Sensing of Environment, 2018, 209: 227

[27] ChenJ, ChenJ, LiaoA P, CaoX, ChenX H, HeC Y, HanG, PengS, LuM, ZhangW W, TongX H, MillsJ. Isprs Journal of Photogrammetry & Remote Sensing, 2015, 103: 7