Leveraging machine learning and deep learning for SAR image classification
Abstract
The study conducted a comprehensive analysis of contemporary machine learning and deep learning methods for the classification of synthetic aperture radar (SAR) images. The primary objective was to identify architectures and approaches that ensure high classification accuracy while optimising computational efficiency. Particular emphasis was placed on addressing key challenges, including speckle noise, geometric distortions, and the limited availability of labelled data. The research methodology involved a systematic review of the scientific literature from 2015 to 2024 and an analysis of the polarisation characteristics of SAR images using the Copernicus Browser platform. The effectiveness of traditional machine learning methods, such as Support Vector Machines and Random Forest, was evaluated alongside modern deep learning architectures, including ResNet, U-Net, and Vision Transformer. Special attention was given to the impact of adaptive speckle noise filtering using the Lee filter with varying window sizes (3 × 3, 5 × 5, and 7 × 7) on classification performance. The results demonstrated that deep neural networks outperform traditional methods due to their ability to automatically extract hierarchical feature representations. ResNet achieved high classification accuracy, U-Net proved effective for segmentation, and Vision Transformer captured global dependencies. The optimal balance between speckle noise suppression and detail preservation was found when applying the Lee filter with a 5 × 5 window size. A persistent challenge remains the limited availability of labelled data. To address this issue, semi-supervised learning was explored, as it enhances feature normalisation and model performance. A promising avenue for further research is the utilisation of complex-valued neural networks to optimise computational costs. The findings of this study have practical significance for the automated classification of SAR images in environmental monitoring, agricultural land assessment, and remote sensing applications
Keywords
neural networks; complex-valued calculations; polarisation characteristics; radar sensing; speckle filtering
References
[1] Attioui, S., & Arivazhagan, S. (2020). Unsupervised change detection method in SAR images based on deep belief network using an improved fuzzy c-means clustering algorithm. IET Image Processing, 15(4), 974-982. doi: 10.1049/ipr2.12078.
[2] Bhattacharjee, S., Shanmugam, P., & Das, S. (2024). A novel lightweight multi-attentive general ship detection model for detection of ships in optical and SAR satellite imagery. In M.F. Carlsohn (Ed.), Real-time processing of image, depth, and video information 2024 (article number 130000B). Strasbourg: Society of Photo-Optical Instrumentation Engineers. doi: 10.1117/12.3016869.
[3] Chen, G., Li, Z., Zhou, Q., & Liu, С. (2024). SAR image despeckling based on gradient domain convolutional sparse coding. In Z. Zhang & C. Li (Eds.), Fifteenth international conference on signal processing systems (ICSPS 2023) (article number 130911W). Xi’an: Society of Photo-Optical Instrumentation Engineers. doi: 10.1117/12.3023324.
[4] Chu, B., Chen, J., Zeng, H., Chen, J., Zhu, J., Wang, M., & Gao, X. (2024). An optical and SAR image registration method based on bidirectional style transfer and hybrid feature descriptor. In Y. Wang & T. Chen (Eds.), 2023 4th international conference on geology, mapping and remote sensing (ICGMRS 2023). Wuhan: Society of PhotoOptical Instrumentation Engineers. doi: 10.1117/12.3019591.
[5] Copernicus Browser. (n.d.). Retrieved from https://browser.dataspace.copernicus.eu/.
[6] Cui, W., Zhang, Y., Zhang, X., Li, L., & Liou, F. (2020). Metal additive manufacturing parts inspection using convolutional neural network. Applied Sciences, 10(2), article number 545. doi: 10.3390/app10020545.
[7] Ding, R. (2023). Which network is stronger? Le Net, Alex Net and VGG on image classification. Applied and Computational Engineering, 4, 294-300. doi: 10.54254/2755-2721/4/20230476.
[8] Evans, B., Faul, A., Fleming, A., Vaughan, D.G., & Hosking, J.S. (2023). Unsupervised machine learning detection of iceberg populations within sea ice from dual-polarisation SAR imagery. Remote Sensing of Environment, 297, article number 113780. doi: 10.1016/j.rse.2023.113780.
[9] Filho, J.F.M.R., & Bélanger, P. (2021). Probe standoff optimization method for phased array ultrasonic TFM imaging of curved parts. Sensors, 21(19), article number 6665. doi: 10.3390/s21196665.
[10 Gavrylenko, S., & Chelfk, V. (2023). Development of method base on fuzzy decision trees for identification of the computer systems state. Navigation and Communication Systems, 1(71), 78-83. doi: 10.26906/sunz.2023.1.078.
[11] Hochstuhl, S., Pfeffer, N., Thiele, A., Hammer, H., & Hinz, S. (2023). Your input matters – comparing real-valued PolSAR data representations for CNN-based segmentation. Remote Sensing, 15(24), article number 5738. doi: 10.3390/rs15245738.
[12] Huang, Z. (2024). A research on image recognition and classification based on traditional machine learning and deep learning. Transactions on Computer Science and Intelligent Systems Research, 5, 766-773. doi: 10.62051/0dbqaa10.
[13] Imad, H., Sara, Z., Hajji, M., Yassine, T., & Abdelkrim, N. (2024). Recent advances in SAR image analysis using deep learning approaches: Examples of speckle denoising and change detection. In B. Benhala, A. Raihani & M. Qbadou (Eds.), 4th international conference on innovative research in applied science, egineering and technology (pp. 1-6). Morocco: Institute of Electrical and Electronics Engineers. doi: 10.1109/IRASET60544.2024.10549456.
[14] Iqbal, J., Vogt, M., & Bajorath, J. (2020). Activity landscape image analysis using convolutional neural networks. Journal of Cheminformatics, 12, article number 34. doi: 10.1186/s13321-020-00436-5.
[15] Kanmani, K., Padmanabhan, V., & Pari, P. (2023). Accuracy assessment of different classifiers for sustainable development in landuse and landcover mapping using sentinel SAR and landsat-8 data. EAI Endorsed Transactions on Energy Web, 10. doi: 10.4108/ew.4141.
[16] Konishi, B., Hirose, A., & Natsuaki, R. (2021). Complex-valued reservoir computing for interferometric SAR applications with low computational cost and high resolution. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14, 7981-7993. doi: 10.1109/jstars.2021.3102620.
[17] Kruk, R., Fuller, M.C., Komarov, A.S., Isleifson, D., & Jeffrey, I. (2020). Proof of concept for sea ice stage of development classification using deep learning. Remote Sensing, 12(15), article number 2486. doi: 10.3390/ rs12152486.
[18] Li, Q., Bai, X., Hu, L., Li, L., Bao, Y., Geng, X., & Yan, X.-H. (2024). SAR image semantic segmentation of typical oceanic and atmospheric phenomena. Earth System Science Data. doi: 10.5194/essd-2024-222.
[19] Li, X., Wang, Z., Li, J., & Mu, P. (2023). Swin-transformer based target detection with enhanced maritime SAR images data. In S. Xu & S. Sazena (Eds.), International conference on signal processing and communication technology (SPCT 2022). Harbin: Society of Photo-Optical Instrumentation Engineers. doi: 10.1117/12.2673920.
[20] Li, Y., Zhang, S., Li, X., & Ye, F. (2023). Remote sensing image classification with few labeled data using semisupervised learning. Wireless Communications and Mobile Computing, 2023(1), article number 7724264. doi: 10.1155/2023/7724264.
[21] Liu, S., Pu, N., Cao, J., & Zhang, K. (2022). Synthetic aperture radar image despeckling based on multi-weighted sparse coding. Entropy, 24(1), article number 96. doi: 10.3390/e24010096.
[22] Liu, T., Li, Y., Cao, Y., & Shen, Q. (2017). Change detection in multitemporal synthetic aperture radar images using dual-channel convolutional neural network. Journal of Applied Remote Sensing, 11(4), article number 042615. doi: 10.1117/1.jrs.11.042615.
[23] Majji, S.R., Chalumuri, A., Kune, R., & Manoj, B.S. (2022). Quantum processing in fusion of SAR and optical images for deep learning: A data-centric approach. IEEE Access, 10, 73743-73757. doi: 10.1109/ access.2022.3189474.
[24] Meng, H., Li, C., Liu, Y., Gong, Y., He, W., & Zou, M. (2023). Corn land extraction based on integrating optical and SAR remote sensing images. Land, 12(2), article number 398. doi: 10.3390/land12020398.
[25] Monsalve-Tellez, J.M., Torres-León, J.L., & Garcés-Gómez, Y.A. (2022). Evaluation of SAR and optical image fusion methods in oil palm crop cover classification using the random forest algorithm. Agriculture, 12(7), article number 955. doi: 10.3390/agriculture12070955.
[26] Nillmani, N., Sharma, N., Saba, L., Khanna, N., Kalra, M., Fouda, M., & Suri, J. (2022). Segmentation-based classification deep learning model embedded with explainable AI for COVID-19 detection in chest X-ray scans. Diagnostics, 12(9), article number 2132. doi: 10.3390/diagnostics12092132.
[27] Oumarou, H., & Rismayanti, N. (2024). Automated classification of empon plants: A comparative study using Hu Moments and K-NN algorithm. Indonesian Journal of Data and Science, 4(3), 206-214. doi: 10.56705/ijodas. v4i3.115.
[28] Pan, J., Zhang, H., Wu, W., Gao, Z., & Wu, W. (2022). Multi-domain integrative Swin transformer network for sparse-view tomographic reconstruction. Patterns, 3(6), article number 100498. doi: 10.1016/j. patter.2022.100498.
[29] Poplavskyi, O. (2024). Information technology for image data processing based on hybrid neural networks using geometric features. Information Technologies and Computer Engineering, 21(2), 4-16. doi: org/10.31649/19999941-2024-60-2-4-16.
[30] Schmitt, M., Hughes, L.H., & Zhu, X.X. (2018). The SEN1-2 dataset for deep learning in SAR-optical data fusion. ISPRS Annals of the Photogrammetry Remote Sensing and Spatial Information Sciences, 4(1), 141-146. doi: 10.5194/isprs-annals-iv-1-141-2018.
[31] Villarroya-Carpio, A., & López-Sánchez, J.M. (2023). Multi-annual evaluation of time series of Sentinel-1 interferometric coherence as a tool for crop monitoring. Sensors, 23(4), article number 1833. doi: 10.3390/ s23041833.
[32] Vu, V.T., Pettersson, M.I., Palm, B.G., Alves, D.I., & Gomes, N.R. (2021). Changing flight heading during pass to enhance SAR change detection performance. IET Radar Sonar & Navigation, 15(8), 817-826. doi: 10.1049/ rsn2.12058.
[33] Wang, L., Jin, G., Xiong, X., Zhang, H., & Wu, K. (2022). Object-based automatic mapping of winter wheat based on temporal phenology patterns derived from multitemporal Sentinel-1 and Sentinel-2 imagery. ISPRS International Journal of Geo-Information, 11(8), article number 424. doi: 10.3390/ijgi11080424.
[34] Wei, B., Huang, M., Zhang, Y., Xu, Y., Liu, X., & Xiang, X. (2021). Boosting ship detection in SAR images with complementary pretraining techniques. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14, 8941-8954. doi: 10.1109/jstars.2021.3109002.
[35] Wu, Y., Xia, Y., Jiang, M., Li, S., Chen, M., Zhao, Y., & Teng, D. (2023). Multisource remote sensing image registration based on geometric constraints. In C. Zuo (Ed.), International conference on remote sensing, surveying, and mapping (RSSM 2023) (article number 127100J). Changsha: Society of Photo-Optical Instrumentation Engineers. doi: 10.1117/12.2682662.
[36] Xue, Z., & Zhang, M. (2020). Multiview low-rank hybrid dilated network for SAR target recognition using limited training samples. IEEE Access, 8, 227847-227856. doi: 10.1109/access.2020.3046274.
[37] Zeng, Z., Tan, X., Chen, Z., Huang, Y., Tang, S., & Wan, J. (2022). Robust image similarity metric method for SAR images. Electronics Letters, 58(13), 508-510. doi: 10.1049/ell2.12516.
[38] Zengguo, S., Zhao, M., & Jia, B. (2021). A GF-3 SAR image dataset of road segmentation. Information Technology and Control, 50(1), 89-101. doi: 10.5755/j01.itc.50.1.27987.
[39] Zhang, H., Li, G., & Lin, H. (2016). An automatic co-registration approach for optical and SAR data in urban areas. Annals of GIS, 22(3), 235-243. doi: 10.1080/19475683.2016.1199595.
[40] Zhang, R., Tang, X., You, S., Duan, K., Xiang, H., & Luo, H. (2020). A novel feature-level fusion framework using optical and SAR remote sensing images for land use/land cover (LULC) classification in cloudy mountainous area. Applied Sciences, 10(8), article number 2928. doi: 10.3390/app10082928.
[41] Zhong, Z., Zheng, L., Kang, G., Li, S., & Yang, Y. (2020). Random erasing data augmentation. Proceedings of the AAAI Conference on Artificial Intelligence, 34(7), 13001-13008. doi: 10.1609/aaai.v34i07.7000.