A method for improving the robustness of neural network model for aerial image matching

Viacheslav Moskalenko, Alona Moskalenko, Yuriy Moskalenko

Abstract


Neural network–based image matching techniques are increasingly employed in aerial image analysis—particularly for UAV navigation, localization, and mapping. However, their sensitivity to structured visual distortions (e.g., shadows, illumination changes, and terrain variability) limits robustness under real-world conditions. Addressing this challenge, we propose a training methodology that enhances both robustness and cross-domain generalization of feature-matching models by integrating adversarial procedural noise with activation‐function modification. During training, structured noise patterns (Perlin, Gabor, and Worley) are synthesized and applied in an adversarial manner, while the standard ReLU activation is replaced by a hybrid LeakyReLU6 to mitigate sensitivity to local perturbations. We evaluate our approach on both detector-based (SuperPoint + SuperGlue) and detector-free (LoFTR) architectures using the Aerial Image Matching Benchmark Dataset and further assess cross-domain performance on the HPatches dataset. Experimental results show that our method yields over 4 % absolute improvements in matching precision and recall on noisy test data for both classes of models. Ablation studies confirm that these gains are attributable to the synergistic effect of procedural noise and LeakyReLU6. Moreover, models trained with our procedure exhibit significantly smaller performance drops when transferred to HPatches, demonstrating enhanced generalization relative to conventionally trained counterparts. To our knowledge, this is the first work to combine adversarial procedural‐noise training with activation‐function constraints for aerial image matching. Beyond improved noise resistance, our method advances cross-domain applicability and is readily extendable to diverse neural‐network architectures.

Keywords


image matching; robustness; adversarial training; procedural noise; aerial image

Full Text:

PDF

References


ISIK, M. Comprehensive empirical evaluation of feature extractors in computer vision. PeerJ Computer Science, 2024, vol. 10, article no. e2415. DOI: 10.7717/peerj-cs.2415.

Verykokou, S., & Ioannidis, C. Image Matching: A Comprehensive Overview of Conventional and Learning-Based Methods. Encyclopedia, 2025, vol. 5, iss. 1, article no. 4. DOI: 10.3390/encyclopedia5010004.

Xu, S., Chen, S., Xu, R., Wang, C., Lu, P., & Guo, L. Local feature matching using deep learning: A survey. Information Fusion, 2024, vol. 107, article no. 102344. DOI: 10.1016/j.inffus.2024.102344.

Bonilla, S., Di Vece, C., Daher, R., Ju, X., Stoyanov, D., Vasconcelos, F., & Bano, S. Mismatched: Evaluating the Limits of Image Matching Approaches and Benchmarks (Version 2). arXiv, 2024. DOI: 10.48550/ARXIV.2408.16445.

Zhang, X., Xiao, F., Zheng, M., & Xie, Z. UAV image matching from handcrafted to deep local features. European Journal of Remote Sensing, 2024, vol. 57, iss. 1. DOI: 10.1080/22797254.2024.2307619.

Khurshid, M., Shahzad, M., Khattak, H. A., Malik, M. I., & Fraz, M. M. Vision-Based 3-D Localization of UAV Using Deep Image Matching. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024, vol. 17, pp. 12020–12030. DOI: 10.1109/jstars.2024.3422310.

Kim, Y., Back, S., Song, D., & Lee, B.-Y. Aerial Map-Based Navigation by Ground Object Pattern Matching. Drones, 2024, vol. 8, iss. 8, article no. 375. DOI: 10.3390/drones8080375.

Śledziowski, J., Terefenko, P., Giza, A., Forczmański, P., Łysko, A., Maćków, W., Stępień, G., Tomczak, A., & Kurylczyk, A. Application of Unmanned Aerial Vehicles and Image Processing Techniques in Monitoring Underwater Coastal Protection Measures. Remote Sensing, 2022, vol. 14, iss. 3, article no. 458. DOI: 10.3390/rs14030458.

Zhang, X., He, Z., Ma, Z., Wang, Z., Wang, L. LLFE: A Novel Learning Local Features Extraction for UAV Navigation Based on Infrared Aerial Image and Satellite Reference Image Matching. Remote Sensing, 2021, vol. 13, iss. 22, article no. 4618. DOI: 10.3390/rs13224618.

Koch, T., Zhuo, X., Reinartz, P., & Fraundorfer, F. A new paradigm for matching UAV- and aerial images. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2016, vol. III–3, pp. 83–90. DOI: 10.5194/isprs-annals-iii-3-83-2016.

Bian, X., He, Z., Gong, Z., & Ren, K. An optimized feature extraction algorithm based on SuperPoint. In C. Zuo (Ed.), AOPC 2022: AI in Optics and Photonics, 2023. 9 p. SPIE. DOI: 10.1117/12.2651943.

Petrakis, G. A SuperPoint neural network implementation for accurate feature extraction in unstructured environments. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2023, vol. XLVIII-1/W2-2023, pp. 1215–1222. DOI: 10.5194/isprs-archives-xlviii-1-w2-2023-1215-2023.

Revaud, J., Weinzaepfel, P., De Souza, C., Pion, N., Csurka, G., Cabon, Y., & Humenberger, M. R2D2: Repeatable and reliable detector and descriptor (Version 2). arXiv, 2019. DOI: 10.48550/ARXIV.1906.06195.

Wang, T., Lin, X., Zhang, P. Enhancing astronomical image stitching with XFeat: A deep learning approach. 2024 International Conference on Image Processing, Computer Vision and Machine Learning (ICICML), 2024, pp. 708–712. IEEE. DOI: 10.1109/icicml63543.2024.10958013.

Li, W. SuperGlue-based deep learning method for image matching from multiple viewpoints. Proceedings of the 2023 8th International Conference on Mathematics and Artificial Intelligence, 2023, pp. 53–58. ACM. DOI: 10.1145/3594300.3594310.

Lindenberger, P., Sarlin, P.-E., & Pollefeys, M. LightGlue: Local feature matching at light speed. 2023 IEEE/CVF International Conference on Computer Vision (ICCV), 2023. IEEE. DOI: 10.1109/iccv51070.2023.01616.

Hoshi, S., Ito, K., & Aoki, T. Accurate and robust image correspondence for structure-from-motion and its application to multi-view stereo. 2022 IEEE International Conference on Image Processing (ICIP), 2022, pp. 2626–2630. IEEE. DOI: 10.1109/icip46576.2022.9897304.

Kumbhar, M. F. Image denoising using autoencoders. In Artificial Intelligence and Knowledge Processing, 2023, pp. 137–144. CRC Press. DOI: 10.1201/9781003328414-13.

Moskalenko, V., Kharchenko, V., Moskalenko, A., & Kuzikov, B. Resilience and resilient systems of artificial intelligence: Taxonomy, models and methods. Algorithms, 2023, vol. 16, iss. 3, article no. 165. DOI: 10.3390/a16030165.

Pogorzelski, T. P. Aerial Image Matching Benchmark Dataset [Dataset]. IEEE DataPort, 2025. DOI: 10.21227/JD5S-AY89.

Sarlin, P.-E., DeTone, D., Malisiewicz, T., Rabinovich, A. SuperGlue: Learning feature matching with graph neural networks (Version 2). arXiv, 2019. DOI: 10.48550/ARXIV.1911.11763.

Bai, T., Luo, J., Zhao, J., Wen, B., & Wang, Q. Recent Advances in Adversarial Training for Adversarial Robustness. Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence (IJCAI-21), 2021, pp. 4312–4321. DOI: 10.24963/ijcai.2021/591.

Athalye, A., Carlini, N., & Wagner, D. Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples. arXiv, 2018. DOI: 10.48550/ARXIV.1802.00420.

Qiu, P., Wang, Q., Wang, D., Lyu, Y., Lu, Z., & Qu, G. Mitigating Adversarial Attacks for Deep Neural Networks by Input Deformation and Augmentation. 2020 25th Asia and South Pacific Design Automation Conference (ASP-DAC), 2020, pp. 157–162. DOI: 10.1109/ASP-DAC47756.2020.9045107.

Zhang, B., Tondi, B., Lv, X., Barni, M. Challenging the Adversarial Robustness of DNNs Based on Error-Correcting Output Codes. Security and Communication Networks, 2020, vol. 2020, pp. 1–11. DOI: 10.1155/2020/8882494.

Maurício, J., Domingues, I., & Bernardino, J. Comparing Vision Transformers and Convolutional Neural Networks for Image Classification: A Literature Review. Applied Sciences, 2023, vol. 13, iss. 9, article no. 5521. DOI: 10.3390/app13095521.

Assran, M., Caron, M., Misra, I., Bojanowski, P., Bordes, F., Vincent, P., Joulin, A., Rabbat, M., & Ballas, N. Masked Siamese Networks for Label-Efficient Learning. Lecture Notes in Computer Science, Springer Nature Switzerland, 2022, pp. 456–473. DOI: 10.1007/978-3-031-19821-2_26.




DOI: https://doi.org/10.32620/reks.2025.3.12

Refbacks

  • There are currently no refbacks.