Geometry-Aware Feature Enhancement With Linear Attention for Robust UAV Photogrammetric Reconstruction Under Weak and Isomorphic Textures
DOI:
https://doi.org/10.5566/ias.3932Keywords:
UAV photogrammetry, Structure-from-Motion, Feature descriptor, Linear Attention, Weak textureAbstract
UAV photogrammetry constitutes a fundamental technology for the lifecycle maintenance and digital twin construction of large-scale transport infrastructure. However, standard Structure-from-Motion (SfM) pipelines frequently falter in scenarios characterized by weak textures, such as asphalt, and repetitive patterns. This deficiency leads to severe feature ambiguity and sparse reconstruction voids in large-scale infrastructure scenes. Furthermore, existing deep learning descriptors typically neglect explicit spatial attributes and suffer from the computational burden of quadratic-complexity attention mechanisms, hindering deployment on edge devices. Building on the FeatureBooster-style descriptor enhancement paradigm, this study adapts a lightweight geometry-aware reconstruction framework to unmanned aerial vehicle infrastructure inspection. The methodology integrates a dual-stream descriptor enhancement model. Following the descriptor-boosting idea of combining local descriptors with geometric keypoint attributes, the adapted model embeds spatial attributes into the feature space to alleviate isomorphic ambiguity. Meanwhile, the modified cross-perception module replaces the original Attention-Free Transformer (AFT) -Simple setting with AFT-Full and incorporates SwiGLU to capture global context efficiently. Experiments on real-world datasets of complex interchanges, urban highways, and campus scenes validate that the adapted descriptor enhancement strategy effectively reduces point cloud voids and reduces reprojection error by approximately 17.3% compared with the original SIFT baseline on Dataset 1. Notably, this study empirically identifies an efficiency compensation phenomenon, wherein superior feature quality accelerates downstream geometric verification and optimization stages. Consequently, although feature enhancement introduces marginal overhead, the overall reconstruction time is reduced in certain datasets. This work provides an application-oriented adaptation and evaluation of FeatureBooster-style descriptor enhancement for geometry-consistent and computationally efficient infrastructure digitalization.
References
Alcantarilla PF, Solutions T (2011). Fast explicit diffusion for accelerated features in nonlinear scale spaces. IEEE Trans Pattern Anal Mach Intell 34(7):1281–98.
Arandjelović R, Zisserman A (2012). Three things everyone should know to improve object retrieval. https://doi.org/10.1109/CVPR.2012.6248018
Bu S, Zhao Y, Wan G, Liu Z (2016). Map2DFusion: Real-time incremental UAV image mosaicing based on monocular SLAM. Proc IEEE/RSJ Int Conf Intell Robots Syst, 4564–71. https://doi.org/10.1109 /iros.2016.7759672
Cakir F, He K, Xia X, Kulis B, Sclaroff S (2019). Deep metric learning to rank. Proc IEEE/CVF Conf Comput Vis Pattern Recogn, 1861–70.
Chen H, Luo Z, Zhou L, Tian Y, Zhen M, Fang T, Quan L (2022). ASpanFormer: Detector-free image matching with adaptive span transformer. Eur Conf Comput Vis, 20–36. Cham: Springer Nature Switzerland.
Chen Y, Liu X, Zhu B, Zhu D, Zuo X, Li Q (2025). UAV image-based 3D reconstruction technology in landslide disasters: A review. Remote Sens 17(17):3117. https://doi.org/10.3390/rs17173117
DeTone D, Malisiewicz T, Rabinovich A (2018). SuperPoint: Self-supervised interest point detection and description. Proc IEEE Conf Comput Vis Pattern Recogn Workshops, 224–36.
Dusmanu M, Miksik O, Schönberger JL, Pollefeys M (2021). Cross-descriptor visual localization and mapping. Proc IEEE/CVF Int Conf Comput Vis, 6058–67.
Han S, Han D (2024). Enhancing direct georeferencing using real-time kinematic UAVs and structure from motion-based photogrammetry for large-scale infrastructure. Drones 8(12):736.
Han K, Wang Y, Chen H, Chen X, Guo J, Liu Z, Tang Y, Xiao A, Xu C, Xu Y, Yang Z, Zhang Y, Tao D (2022). A survey on vision transformer. IEEE Trans Pattern Anal Mach Intell 45(1):1–1. https://doi.org/10.1109/ tpami.2022.3152247
He L, Wang X, Zhang H (2016). M2DP: A novel 3D point cloud descriptor and its application in loop closure detection. Proc IEEE/RSJ Int Conf Intell Robots Syst. https://doi.org/10.1109/iros.2016.7759060
Hornik K, Stinchcombe M, White H (1989). Multilayer feedforward networks are universal approximators. Neural Netw 2(5):359–66. https://doi.org/10.1016/ 0893-6080(89)90020-8
Ji S, Zeng C, Zhang Y, Duan Y (2023). An evaluation of conventional and deep learning-based image-matching methods on diverse datasets. Photogramm Rec 38(182):137–59. https://doi.org/10.1111/phor.12445
Jiang S, Jiang C, Jiang W (2020). Efficient structure from motion for large-scale UAV images: A review and a comparison of SfM tools. ISPRS J Photogramm Remote Sens, 167, 230–51. https://doi.org/10.1016/ j.isprsjprs.2020.04.016
Ke S, Yang K, Zhan C, Liao S, Ma Y, Shang H, Shen E, Li Z, Zhang Z, Chen Z (2025). Advances in earth observation using unmanned aerial vehicles: A bibliometric and content analysis, 2000–2024. Geocarto Int, 40(1). https://doi.org/10.1080/ 10106049.2025.2600779
Kerbl B, Kopanas G, Leimkühler T, Drettakis G (2023). 3D Gaussian splatting for real-time radiance field rendering. ACM Trans Graph 42(4):1–14. https://doi.org/10.1145/3592433
Koohmishi M, Kaewunruen S, Chang L, Guo Y (2024). Advancing railway track health monitoring: Integrating GPR, InSAR and machine learning for enhanced asset management. Autom Constr, 162, 105378. https://doi.org/10.1016/j.autcon.2024.105378
Lee E, Park S, Jang H, Choi W, Sohn HG (2024). Enhancement of low-cost UAV-based photogrammetric point cloud using MMS point cloud and oblique images for 3D urban reconstruction. Measurement, 226, 114158.
Lindenberger P, Sarlin PE, Pollefeys M (2023). LightGlue: Local feature matching at light speed. Proc IEEE/CVF Int Conf Comput Vis, 17627–38.
Lowe DG (2004). Distinctive image features from scale-invariant keypoints. Int J Comput Vis 60(2):91–110. https://doi.org/10.1023/B:VISI.0000029664.99615.94
Morel JM, Yu G (2009). ASIFT: A new framework for fully affine invariant image comparison. SIAM J Imaging Sci 2(2):438–69. https://doi.org/10.1137/ 080732730
Moussa LG, Diaconu R, Watt MS, Muñoz E, Casado MR, Broadbent EN, Bruscolini M, Doaemo W, Mohan M (2024). UAVs as a tool for optimizing boat-supported flood evacuation operations. Drones 8(11):621. https://doi.org/10.3390/drones8110621
Rublee E, Rabaud V, Konolige K, Bradski G (2011). ORB: An efficient alternative to SIFT or SURF. https://doi.org/10.1109/ICCV.2011.6126544
Schönberger JL, Frahm JM (2016). Structure-from-Motion revisited. Proc IEEE Conf Comput Vis Pattern Recogn. https://doi.org/10.1109/CVPR.2016.445
Shazeer N (2020). GLU variants improve transformer. arXiv. https://doi.org/10.48550/arXiv.2002.05202
Simantiris G, Panagiotakis C (2024). Unsupervised color-based flood segmentation in UAV imagery. Remote Sens 16(12):2126. https://doi.org/10.3390/rs16122126
Sun J, Shen Z, Wang Y, Bao H, Zhou X (2021). LoFTR: Detector-free local feature matching with transformers. Proc IEEE/CVF Conf Comput Vis Pattern Recogn, 8922–31.
Sun J, Yuan G, Song L, Zhang H (2024). Unmanned aerial vehicles in landslide investigation and monitoring: A review. Drones 8(1):30. https://doi.org/10.3390 /drones8010030
Tian Y, Yu X, Fan B, Wu F, Heijnen H, Balntas V (2019). SOSNet: Second order similarity regularization for local descriptor learning. Proc IEEE/CVF Conf Comput Vis Pattern Recogn, 11016–25.
Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones LJ, Gomez AN, Kaiser Ł, Polosukhin I (2017). Attention is all you need. Adv Neural Inf Process Syst, 30.
Wang Q, Zhang J, Yang K, Peng K, Stiefelhagen R (2022). MatchFormer: Interleaving attention in transformers for feature matching. Proc Asian Conf Comput Vis, 2746–62.
Wang X, Liu Z, Hu Y, Xi W, Yu W, Zou D (2023). FeatureBooster: Boosting feature descriptors with a lightweight neural network. Proc IEEE/CVF Conf Comput Vis Pattern Recogn, 7630–39.
Westoby MJ, Brasington J, Glasser NF, Hambrey MJ, Reynolds JM (2012). Structure-from-Motion photogrammetry: A low-cost, effective tool for geoscience applications. Geomorphology, 179, 300–14.
Wu D, Zheng A, Yu W, Cao H, Ling Q, Liu J, Zhou D (2025). Digital twin technology in transportation infrastructure: A comprehensive survey of current applications, challenges, and future directions. Appl Sci 15(4):1911. https://doi.org/10.3390/app15041911
Yan B, Yang F, Qiu S, Wang J, Cai B, Wang S, Hu W (2023). Digital twin in transportation infrastructure management: A systematic review. Intell Transp Infrastruct, 2, liad024.
Yao Y, Luo Z, Li S, Fang T, Quan L (2018). MVSNet: Depth inference for unstructured multi-view stereo. Proc Eur Conf Comput Vis, 767–83.
Yi KM, Trulls E, Lepetit V, Fua P (2016). LIFT: Learned invariant feature transform. Comput Vis ECCV, 467–83. https://doi.org/10.1007/978-3-319-46466-4_28
Zhang Y, Xue Y, Lan C, Xing X, Pang Y, Xu M (2025). Multisource oblique remote sensing image matching with affine-invariant features and geometric constraints. Int J Digit Earth, 19(1). https://doi.org/ 10.1080/17538947.2025.2600881
Downloads
Published
Data Availability Statement
The public dataset (NPU Campus) used in this study is available from the original source cited in the article. The self-collected datasets generated during the current study are available from the corresponding author on reasonable request.
Issue
Section
License
Copyright (c) 2026 Tian Yang Luo, Xiaocong Jiang
This work is licensed under a Creative Commons Attribution 4.0 International License.
How to Cite
