Land Use and Land Cover Mapping Using Deep Learning Based Segmentation Approaches and VHR Worldview-3 Images
<p>(<b>a</b>) General view of the study area and its surroundings. (<b>b</b>) The administrative boundary of Bursa province. (<b>c</b>) The administrative boundaries of Aksu and Kestel sites used in the research.</p> "> Figure 2
<p>Sample patches from the WV-3 images of the study areas, representing different LULC classes adapted from CORINE second-level nomenclature. (<b>a</b>) Heterogeneous agricultural areas, (<b>b</b>) Arable land, (<b>c</b>) Industrial units, (<b>d</b>) Forest, (<b>e</b>) Permanent crops, (<b>f</b>) Inland waters, (<b>g</b>) Continuous urban fabric, (<b>h</b>) Discontinuous urban fabric, (<b>i</b>) Road and rail networks and associated land, (<b>j</b>) Shrub and/or herbaceous vegetation associations, (<b>k</b>) Artificial, non-agricultural vegetated areas, (<b>l</b>) Mine, dump, and construction sites.</p> "> Figure 3
<p>LULC classes and their class-wise distributions (<b>a</b>) Class legend, (<b>b</b>) Aksu dataset, (<b>c</b>) Kestel dataset.</p> "> Figure 4
<p>Sample image patches and their corresponding ground truth masks. (<b>a</b>) sample patches from the Aksu region, (<b>b</b>) sample patches from the Kestel region.</p> "> Figure 5
<p>Flowchart of the used deep neural network architecture which follows an encoder–decoder structure with Atrous convolutions that bypasses the low-level features to the decoder.</p> "> Figure 6
<p>Comparison of visual results of predictions from different encoders. (<b>a</b>) Input images, (<b>b</b>) Ground truth data, (<b>c</b>) Resnext50_32x4d results, (<b>d</b>) Resnet50 results, (<b>e</b>) DPN-68 results, (<b>f</b>) Mobilenetv2 results, and (<b>g</b>) Efficientnet-B2 results.</p> "> Figure 7
<p>Qualitative results of the classifier trained on the Aksu dataset. (<b>a1</b>–<b>a6</b>) show original image patches; (<b>b1</b>–<b>b6</b>) illustrate the corresponding Ground Truth masks, and (<b>c1</b>–<b>c6</b>) show the prediction results with the proposed model.</p> "> Figure 8
<p>Qualitative results of the classifier trained on the Kestel dataset. (<b>a1</b>–<b>a6</b>) show original image patches; (<b>b1</b>–<b>b6</b>) illustrate the corresponding Ground Truth masks, and (<b>c1</b>–<b>c6</b>) show the prediction results with the proposed model.</p> "> Figure 9
<p>Qualitative results of the classifier trained on the Aksu + Kestel dataset. (<b>a1</b>–<b>a6</b>) show original image patches; (<b>b1</b>–<b>b6</b>) illustrate the corresponding Ground Truth masks, and (<b>c1</b>–<b>c6</b>) show the prediction results with the proposed model.</p> ">
Abstract
:1. Introduction
2. Study Area and Dataset
2.1. Study Area and Image Dataset Descriptions
2.2. Dataset Generation
- The Aksu dataset consists of nine categories:
- Discontinuous urban fabric,
- Road and rail networks and associated land,
- Mine, dump, and construction sites,
- Artificial, non-agricultural vegetated areas,
- Arable land,
- Permanent crops,
- Heterogeneous agricultural areas,
- Forest, and
- Inland waters.
- Industrial or commercial units,
- Shrub and/or herbaceous vegetation associations, and
- Continuous urban fabric.
3. Methodological Approach and Experimental Setup
3.1. Implementation Details
3.2. Evaluation Metrics
3.3. Results and Discussion
- Next generation ResNet (ResNeXt), resnext50_32x4d version with 22 M parameters and ImageNet weights,
- Detail-Preserving Network (DPN), DPN68 version with 11 M parameters and ImageNet weights
- EfficientNet, efficientnet-b0, efficientnet-b1, and efficientnet-b2 versions with 4M, 6M, and 7M parameters, respectively and having ImageNet weights.
- MobileNet, mobilenet_v2 version with 2M parameters and ImageNet weights.
Architecture | Parameters | IoU | F1 Score | Precision | Recall |
---|---|---|---|---|---|
ResNeXt50 | 22M | 89.46 | 94.34 | 94.25 | 94.49 |
ResNet50 | 23M | 87.32 | 93.08 | 92.99 | 93.16 |
DPN68 | 11M | 80.83 | 88.61 | 88.61 | 88.61 |
MobileNet v2 | 2M | 79.07 | 88.09 | 88.15 | 88.02 |
Efficientnet-b0 | 4M | 79.94 | 88.48 | 88.42 | 88.55 |
Efficientnet-b1 | 6M | 82.64 | 90.24 | 90.16 | 90.32 |
Efficientnet-b2 | 7M | 83.36 | 90.58 | 90.52 | 90.64 |
4. Conclusions
Author Contributions
Funding
Data Availability Statement
Acknowledgments
Conflicts of Interest
References
- Sertel, E.; Topaloğlu, R.H.; Şallı, B.; Yay Algan, I.; Aksu, G.A. Comparison of landscape metrics for three different level land cover/land use maps. ISPRS Int. J. Geo-Inf. 2018, 7, 408. [Google Scholar] [CrossRef]
- Liu, Y.; Fan, B.; Wang, L.; Bai, J.; Xiang, S.; Pan, C. Semantic labeling in very high resolution images via a self-cascaded convolutional neural network. ISPRS J. Photogramm. Remote Sens. 2018, 145, 78–95. [Google Scholar] [CrossRef]
- Yuan, X.; Shi, J.; Gu, L. A review of deep learning methods for semantic segmentation of remote sensing imagery. Expert Syst. Appl. 2021, 169, 114417. [Google Scholar] [CrossRef]
- Xue, Z.; Li, J.; Cheng, L.; Du, P. Spectral–spatial classification of hyperspectral data via morphological component analysis-based image separation. IEEE Trans. Geosci. Remote Sens. 2015, 53, 70–84. [Google Scholar] [CrossRef]
- Ekim, B.; Sertel, E. Deep neural network ensembles for remote sensing land cover and land use classification. Int. J. Digit. Earth 2021, 14, 1868–1881. [Google Scholar] [CrossRef]
- Ettehadi Osgouei, P.; Sertel, E.; Kabadayı, M.E. Integrated usage of historical geospatial data and modern satellite images reveal long-term land use/cover changes in Bursa/Turkey, 1858–2020. Sci. Rep. 2022, 12, 9077. [Google Scholar] [CrossRef]
- Blaschke, T.; Hay, G.J.; Kelly, M.; Lang, S.; Hofmann, P.; Addink, E.; Queiroz Feitosa, R.; van der Meer, F.; van der Werff, H.; van Coillie, F.; et al. Geographic object-based image analysis—Towards a new paradigm. ISPRS J. Photogramm. Remote Sens. 2014, 87, 180–191. [Google Scholar] [CrossRef]
- Saha, S.; Bovolo, F.; Bruzzone, L. Unsupervised deep change vector analysis for multiple-change detection in VHR images. IEEE Trans. Geosci. Remote Sens. 2019, 57, 3677–3693. [Google Scholar] [CrossRef]
- Castillo-Navarro, J.; Le Saux, B.; Boulch, A.; Audebert, N.; Lefèvre, S. Semi-supervised semantic segmentation in Earth observation: The minifrance suite, dataset analysis and multi-task network study. Mach. Learn. 2021, 1–36. [Google Scholar] [CrossRef]
- Blaschke, T. Object based image analysis for remote sensing. ISPRS J. Photogramm. Remote Sens. 2010, 65, 2–16. [Google Scholar] [CrossRef] [Green Version]
- Qin, R.; Liu, T. A review of landcover classification with very-high resolution remotely sensed optical images—Analysis unit, model scalability and transferability. Remote Sens. 2022, 14, 646. [Google Scholar] [CrossRef]
- Topaloğlu, R.H.; Aksu, G.A.; Ghale, Y.A.G.; Sertel, E. High-resolution land use and land cover change analysis using GEOBIA and landscape metrics: A case of Istanbul, Turkey. Geocarto Int. 2021, 1–27. [Google Scholar] [CrossRef]
- Zhang, X.; Chen, G.; Wang, W.; Wang, Q.; Dai, F. Object-based land-cover supervised classification for very-high-resolution UAV images using stacked denoising autoencoders. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2017, 10, 3373–3385. [Google Scholar] [CrossRef]
- de Pinho, C.M.D.; Fonseca, L.M.G.; Korting, T.S.; de Almeida, C.M.; Kux, H.J.H. Land-cover classification of an intra-urban environment using high-resolution images and object-based image analysis. Int. J. Remote Sens. 2012, 33, 5973–5995. [Google Scholar] [CrossRef]
- Chen, G.; Weng, Q.; Hay, G.J.; He, Y. Geographic object-based image analysis (GEOBIA): Emerging trends and future opportunities. GIScience Remote Sens. 2018, 55, 159–182. [Google Scholar] [CrossRef]
- Neupane, B.; Horanont, T.; Aryal, J. Deep learning-based semantic segmentation of urban features in satellite images: A review and meta-analysis. Remote Sens. 2021, 13, 808. [Google Scholar] [CrossRef]
- Merciol, F.; Faucqueur, L.; Damodaran, B.B.; Rémy, P.-Y.; Desclée, B.; Dazin, F.; Lefèvre, S.; Masse, A.; Sannier, C. GEOBIA at the terapixel scale: Toward efficient mapping of small woody features from heterogeneous VHR scenes. ISPRS Int. J. Geo-Inf. 2019, 8, 46. [Google Scholar] [CrossRef]
- Long, J.; Shelhamer, E.; Darrell, T. Fully Convolutional Networks for Semantic Segmentation. arXiv 2015, arXiv:1411.4038. [Google Scholar]
- Rottensteiner, F.; Sohn, G.; Jung, J.; Gerke, M.; Baillard, C.; Benitez, S.; Breitkopf, U. The ISPRS benchmark on urban object classification and 3D building reconstruction. ISPRS Ann. Photogramm. Remote Sens. Spat. Inf. Sci. 2012, I–3, 293–298. [Google Scholar] [CrossRef]
- ISPRS Potsdam 2D Semantic Labeling—Potsdam. 2019. Available online: https://www2.isprs.org/commissions/comm2/wg4/benchmark/2d-sem-label-potsdam/ (accessed on 1 February 2022).
- ISPRS Vaihingen 2D Semantic Label.—Vaihingen. 2019. Available online: https://www2.isprs.org/commissions/comm2/wg4/benchmark/2d-sem-label-vaihingen/ (accessed on 1 February 2022).
- Comp. Intelligence Group Hyperspectral Remote Sensing Scenes—Grupo de Inteligencia Computacional (GIC). 2019. Available online: http://www.ehu.eus/ccwintco/index.php?title=Hyperspectral_Remote_Sensing_Scenes#Pavia_Centre_and_University (accessed on 1 February 2022).
- Mnih, V. Mnih Massachusetts Building Dataset. Ph.D. Thesis, University of Toronto, Toronto, ON, Canada, 2013. [Google Scholar]
- Papadomanolaki, M.; Vakalopoulou, M.; Karantzalos, K. A Novel Object-based deep learning framework for semantic segmentation of very high-resolution remote sensing data: Comparison with convolutional and fully convolutional networks. Remote Sens. 2019, 11, 684. [Google Scholar] [CrossRef]
- Kemker, R.; Salvaggio, C.; Kanan, C. Algorithms for semantic segmentation of multispectral remote sensing imagery using deep learning. ISPRS J. Photogramm. Remote Sens. 2018, 145, 60–77. [Google Scholar] [CrossRef] [Green Version]
- Audebert, N.; Le Saux, B.; Lefèvre, S. Beyond RGB: Very high resolution urban remote sensing with multimodal deep networks. ISPRS J. Photogramm. Remote Sens. 2018, 140, 20–32. [Google Scholar] [CrossRef]
- Längkvist, M.; Kiselev, A.; Alirezaie, M.; Loutfi, A. Classification and segmentation of satellite orthoimagery using convolutional neural networks. Remote Sens. 2016, 8, 329. [Google Scholar] [CrossRef]
- Fu, G.; Liu, C.; Zhou, R.; Sun, T.; Zhang, Q. Classification for high resolution remote sensing imagery using a fully convolutional network. Remote Sens. 2017, 9, 498. [Google Scholar] [CrossRef]
- Zhang, Y. A new automatic approach for effectively fusing landsat 7 as well as IKONOS images. IEEE Int. Geosci. Remote Sens. Symp. 2002, 4, 2429–2431. [Google Scholar] [CrossRef]
- Wang, P.; Sertel, E. Channel–spatial attention-based pan-sharpening of very high-resolution satellite images. Knowl. Based Syst. 2021, 229, 107324. [Google Scholar] [CrossRef]
- Sklearn Package. Available online: https://scikit-learn.org/stable/about.html#citing-scikit-learn (accessed on 1 February 2022).
- Ekim, B.; Sertel, E.; Kabadayı, M.E. Automatic Road extraction from historical maps using deep learning techniques: A regional case study of Turkey in a German World War II Map. Int. J. Geo-Inf. 2021, 10, 492. [Google Scholar] [CrossRef]
- Ronneberger, O.; Fischer, P.; Brox, T. U-Net: Convolutional networks for biomedical image segmentation. In Proceedings of the Medical Image Computing and Computer-Assisted Intervention—MICCAI 2015, Munich, Germany, 5–9 October 2015; Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F., Eds.; Springer International Publishing: Cham, Switzerland, 2015; pp. 234–241. [Google Scholar] [CrossRef]
- Avci, C.; Sertel, E.; Kabadayı, M.E. Deep Learning Based Road Extraction from Historical Maps. IEEE Geosci. Remote Sens. Lett. 2022, PP, 1. [Google Scholar] [CrossRef]
- Chen, L.-C.; Papandreou, G.; Schroff, F.; Adam, H. Rethinking atrous convolution for semantic image segmentation. arXiv 2017, arXiv:1706.05587. [Google Scholar]
- Xie, S.; Girshick, R.; Dollár, P.; Tu, Z.; He, K. Aggregated residual transformations for deep neural networks. In Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 21–26 July 2017; pp. 5987–5995. [Google Scholar] [CrossRef]
- Sudre, C.H.; Li, W.; Vercauteren, T.; Ourselin, S.; Cardoso, M.J. Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations. arXiv 2017, arXiv:1707.03237. [Google Scholar] [CrossRef] [Green Version]
- Mulyanto, M.; Faisal, M.; Prakosa, S.W.; Leu, J.-S. Effectiveness of focal loss for minority classification in network intrusion detection systems. Symmetry 2021, 13, 4. [Google Scholar] [CrossRef]
- Congalton, R.G. A review of assessing the accuracy of classifications of remotely sensed data. Remote Sens. Environ. 1991, 37, 35–46. [Google Scholar] [CrossRef]
- Chen, L.-C.; Zhu, Y.; Papandreou, G.; Schroff, F.; Adam, H. Encoder-decoder with atrous separable convolution for semantic image segmentation. In Computer Vision—ECCV; Ferrari, V., Hebert, M., Sminchisescu, C., Weiss, Y., Eds.; Springer International Publishing: Cham, Switzerland, 2018; Volume 11211, pp. 833–851. ISBN 978-3-030-01233-5. [Google Scholar]
- Zhao, H.; Shi, J.; Qi, X.; Wang, X.; Jia, J. Pyramid scene parsing network. arXiv 2017, arXiv:1612.01105. [Google Scholar]
- Li, H.; Xiong, P.; An, J.; Wang, L. Pyramid attention network for semantic segmentation. arXiv 2018, arXiv:1805.10180. [Google Scholar]
- Zhou, Z.; Siddiquee, M.M.R.; Tajbakhsh, N.; Liang, J. UNet++: A nested U-Net architecture for medical image segmentation. arXiv 2018, arXiv:1807.10165. [Google Scholar]
- Lin, T.-Y.; Dollár, P.; Girshick, R.; He, K.; Hariharan, B.; Belongie, S. Feature pyramid networks for object detection. arXiv 2017, arXiv:1612.03144. [Google Scholar]
- Chaurasia, A.; Culurciello, E. LinkNet: Exploiting encoder representations for efficient semantic segmentation. In Proceedings of the 2017 IEEE Visual Communications and Image Processing (VCIP), St. Petersburg, FL, USA, 10–13 December 2017; pp. 1–4. [Google Scholar] [CrossRef]
- Ghosh, S.; Huo, M.; Shawkat, M.S.A.; McCalla, S. Using convolutional encoder networks to determine the optimal magnetic resonance image for the automatic segmentation of Multiple Sclerosis. Appl. Sci. 2021, 11, 8335. [Google Scholar] [CrossRef]
- Zhang, P.; Ke, Y.; Zhang, Z.; Wang, M.; Li, P.; Zhang, S. Urban land use and land cover classification using novel deep learning models based on high spatial resolution satellite imagery. Sensors 2018, 18, 3717. [Google Scholar] [CrossRef]
- Bengana, N.; Heikkila, J. Improving land cover segmentation across satellites using domain adaptation. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2021, 14, 1399–1410. [Google Scholar] [CrossRef]
Dataset | Number of Classes | Number of Patches | Number of Patches in Train/ Validation/Test Sets |
---|---|---|---|
Aksu | 10 | 599 | 419/120/60 |
Kestel | 13 | 265 | 185/53/27 |
Aksu + Kestel | 13 | 784 | 549/157/78 |
Architecture | IoU | F-1 Score | Precision | Recall |
---|---|---|---|---|
DeepLabv3+ | 89.46 | 94.35 | 94.25 | 94.49 |
PAN | 82.78 | 90.37 | 90.34 | 90.47 |
U-Net++ | 81.54 | 89.54 | 89.63 | 89.45 |
FPN | 76.45 | 86.39 | 86.39 | 86.38 |
Linknet | 74.75 | 84.99 | 84.95 | 85.04 |
PSPNet | 71.20 | 82.44 | 82.44 | 82.45 |
Dataset | IoU | F1 Score | Precision | Recall |
---|---|---|---|---|
Aksu | 89.46 | 94.35 | 94.25 | 94.49 |
Kestel | 81.64 | 89.65 | 89.76 | 89.54 |
Aksu + Kestel | 86.92 | 92.85 | 92.84 | 92.86 |
Aksu | Kestel | Aksu + Kestel | |
---|---|---|---|
Forest | 0.952 | 0.918 | 0.968 |
Mine, dump, and construction sites | 0.866 | 0.960 | 0.903 |
Road and rail | 0.612 | 0.683 | 0.779 |
Discontinuous urban fabric | 0.894 | 0.794 | 0.847 |
Arable land | 0.932 | 0.844 | 0.867 |
Heterogeneous agricultural areas | 0.943 | 0.931 | 0.919 |
Permanent crops | 0.908 | 0.917 | 0.850 |
Inland waters | 0.983 | 0.809 | 0.965 |
Artificial, non-agricultural vegetated areas | 0.989 | 0.715 | 0.671 |
Industrial or commercial units | - | 0.967 | 0.954 |
Shrub and/or herbaceous vegetation | - | 0.250 | 0.775 |
Continuous urban fabric | - | 0.986 | 0.983 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Sertel, E.; Ekim, B.; Ettehadi Osgouei, P.; Kabadayi, M.E. Land Use and Land Cover Mapping Using Deep Learning Based Segmentation Approaches and VHR Worldview-3 Images. Remote Sens. 2022, 14, 4558. https://doi.org/10.3390/rs14184558
Sertel E, Ekim B, Ettehadi Osgouei P, Kabadayi ME. Land Use and Land Cover Mapping Using Deep Learning Based Segmentation Approaches and VHR Worldview-3 Images. Remote Sensing. 2022; 14(18):4558. https://doi.org/10.3390/rs14184558
Chicago/Turabian StyleSertel, Elif, Burak Ekim, Paria Ettehadi Osgouei, and M. Erdem Kabadayi. 2022. "Land Use and Land Cover Mapping Using Deep Learning Based Segmentation Approaches and VHR Worldview-3 Images" Remote Sensing 14, no. 18: 4558. https://doi.org/10.3390/rs14184558
APA StyleSertel, E., Ekim, B., Ettehadi Osgouei, P., & Kabadayi, M. E. (2022). Land Use and Land Cover Mapping Using Deep Learning Based Segmentation Approaches and VHR Worldview-3 Images. Remote Sensing, 14(18), 4558. https://doi.org/10.3390/rs14184558