Skip to content
KPI Science News
ISSN 2617-5509 · e‑ISSN 2663-7472 Open Access · CC BY 4.0
Vol. 144 · Issue 3 · 2026 Sep 29, 2026 article

TWO-STAGE SEGMENTATION PIPELINE FOR SATELLITE PARKING LOT PAVED AREA ESTIMATION

OH
Oleksandr Honcharuk Corresponding Department of mathematical methods of system analysis, Educational and Research Institute for Applied System Analysis, National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", Kyiv, Ukraine Ukraine
NN
Nadezhda Nedashkovskaya Department of mathematical methods of system analysis, Educational and Research Institute for Applied System Analysis, National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", Kyiv, Ukraine Ukraine
PublishedSep 29, 2026
PDF (Ukrainian)
KPISN 3 VOL 144 · 3
VOL 144 · ISS 3 · 2026 View issue

Abstract

Background. Accurately estimating paved parking areas is vital for urban planning and environmental monitoring. However, spectral similarities between parking lot and road asphalt make single-stage segmentation unreliable. Existing methods often depend on manually corrected labels. This study introduces a novel pipeline combining free crowd-sourced annotations for coarse localization with human annotations for detailed surface classification. Objective. The aim was to build a two-stage segmentation pipeline wherein the first stage involves localising with boundaries at a coarse level while the second stage is dedicated to segmenting at a fine-grained level of non-paved regions with utilising different sources of supervision for both stages. Methods. Stage 1 uses a U-Net with EfficientNet-B3 trained on 3019 raw OpenStreetMap polygons from 34 US states for boundary detection. Stage 2 masks the image to the detected boundary and trains a second U-Net on 1144 human-annotated masks to segment non-paved regions. Paved area equals boundary area minus non-paved area. Both stages use Dice loss, evaluated by Binary IoU (IoU) at 512x512. Results. Stage 1 delivers an IoU of 0.7886 even without manual correction of the labels. Stage 2 delivers an IoU of 0.6147 and Dice coefficient of 0.8396 for segmentation of the non-paved regions. The difference between the metrics confirms that errors are mainly at boundaries rather than misclassifications on a large scale. The whole pipeline effectively isolates parking lots from their adjacent roads and is also able to locate major non-paved regions. Conclusions. Compared with Qiam et al. [10], we reduced human annotation by 89% and model size by ~18x. The two-stage design combining free OSM labels for coarse localisation with human annotations for fine-grained classification proves viable for paved area estimation. The paradigm extends to related problems such as building material typing and road surface mapping.

Keywords

#semantic segmentation #satellite imagery #parking lot #neural network #U-Net

References

  1. Batmunkh, Ariuntuya. 2024. “Deep Learning for Impervious Surface Segmentation from High Resolution Aerial Imagery.” DOI: https://doi.org/10.32469/10355/106121
  2. Dice, Lee R. 1945. “Measures of the Amount of Ecologic Association Between Species.” Ecology 26 (3): 297–302. DOI: https://doi.org/10.2307/1932409 .
  3. Henry, C., J. Hellekes, N. Merkle, S. M. Azimi, and F. Kurz. 2021. “Citywide Estimation of Parking Space Using Aerial Imagery and OSM Data Fusion with Deep Learning and Fine-Grained Annotation.” The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences XLIII-B2-2021: 479–485. DOI: https://doi.org/10.5194/isprs-archives-XLIII-B2-2021-479-2021 .
  4. Honcharuk, Oleksandr P., and Nadiia I. Nedashkovskaya. 2026. “OpenStreetMap as Free Training Labels for Satellite Parking Lot Segmentation.” In Proc. XXIV All-Ukrainian Scientific-Practical Conference “Theoretical and Applied Problems of Physics, Mathematics and Informatics”. Kyiv, Ukraine: Igor Sikorsky Kyiv Polytechnic Institute, Politekhnika.
  5. URL: https://ela.kpi.ua/handle/123456789/81220
  6. Kaiser, Pascal, Jan Dirk Wegner, Aurelien Lucchi, Martin Jaggi, Thomas Hofmann, and Konrad Schindler. 2017. “Learning Aerial Image Segmentation from Online Maps.” IEEE Transactions on Geoscience and Remote Sensing 55 (11): 6054-6068
  7. DOI: https://doi.org/10.1109/TGRS.2017.2719738 .
  8. URL: https://scispace.com/pdf/learning-aerial-image-segmentation-from-online-
  9. maps-4q145uvzjb.pdf
  10. Kingma, Diederik P., and Jimmy Ba. 2015. “Adam: A Method for Stochastic Optimization.” In International Conference on Learning Representations (Iclr). URL: https://arxiv.org/abs/1412.6980; https://arxiv.org/pdf/1412.6980
  11. Li, Man, Renru Wang, Ana Dai, Weitao Yuan, Guangbin Yang, Lijun Xie, Weili Zhao, and Linglin Zhao. 2025. “Cascade DeepLab Net: A Method for Accurate Extraction of Fragmented Cultivated Land in Mountainous Areas Based on a Cascaded Network.” Agriculture 15 (3): 348. DOI: https://doi.org/10.3390/agriculture15030348 .
  12. Olbricht, Roland. 2024. “Overpass API.” URL: https://overpass-api.de/ .
  13. OpenStreetMap contributors. 2024. “OpenStreetMap.” URL: https://www.openstreetmap.org .
  14. Qiam, Shirin, Saipraneeth Devunuri, and Lewis J. Lehe. 2025. “A Pipeline and NIR-Enhanced Dataset for Parking Lot Segmentation.” In 2025 Ieee/Cvf Winter Conference on Applications of Computer Vision (Wacv), 1227–1236. DOI: https://doi.org/10.1109/WACV61041.2025.00127 .
  15. Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. 2015. “U-Net: Convolutional Networks for Biomedical Image Segmentation.” In Medical Image Computing and Computer-Assisted Intervention – Miccai 2015, 9351:234–241. Cham: Springer International Publishing. DOI: https://doi.org/10.1007/978-3-319-24574-4_28 .
  16. Tan, Mingxing, and Quoc V. Le. 2019. “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks.” URL: https://arxiv.org/abs/1905.11946 . DOI: https://doi.org/10.48550/arXiv.1905.11946
  17. Usmani, Munazza, Francesca Bovolo, and Maurizio Napolitano. 2023. “Remote Sensing and Deep Learning to Understand Noisy Openstreetmap.” Remote Sensing 15 (18): 4639. DOI: https://doi.org/10.3390/rs15184639 .
  18. Vargas-Munoz, John E., Shivangi Srivastava, Devis Tuia, and Alexandre X. Falcão. 2020. “OpenStreetMap: Challenges and Opportunities in Machine Learning and Remote Sensing.” IEEE Geoscience and Remote Sensing Magazine 9 (1): 184–199. DOI: https://doi.org/10.1109/MGRS.2020.2994107 .
  19. Zhong, Qi, Jun Luo, Jingxin Fang, and Yi Chen. 2025. “Two-Stage Cascaded Vision Transformer with Spatial Attention for Dense Settlement Detection in Remote Sensing Imagery.” Npj Heritage Science 13 (1). DOI: https://doi.org/10.1038/s40494-025-01787-8 .
  20. Zgurovsky, Michael, Kostiantyn Yefremov, Sergii Gapon, and Ivan Pyshnograiev. 2023. “Assessment of the Economical Dimension of Sustainable Development of the Ukraine's Regions Based on the Brightness of Night Lights.” System Research and Information Technologies, no. 2: 49–62. DOI: https://doi.org/10.20535/SRIT.2308-8893.2023.2.04 .
  21. Sydorskyi, V. S. 2022. “The Method of Transformation of Image Classification Labels into Segmentation Masks.” Microsystems, Electronics and Acoustics 27 (2): 262933-1–262933-9. DOI: https://doi.org/10.20535/2523-4455.mea.262933 .
§ 06 — Related

Similar articles in this journal

Related peer-reviewed studies published in this journal.
View all issues

AN APPROACH TO CONSTRUCTING AN INFORMATIVE FEATURE SPACE OF TECHNICAL INDICATORS FOR FINANCIAL PROCESSES BASED ON NEURAL NETWORK MODELS

Roman Tertychnyi, Denys Chuprin, Petro Bidyuk

Background. Financial time series are characterized by nonlinearity, non-stationarity, and a high level of noise, which significantly complicates their modeling and…

#informative feature space construction approach #technical indicators #Parkinson volatility #neural networks
10.20535/kpisn.2026.3.359610

Most read articles by the same author(s)

Similar Articles

1-10 of 38

You may also start an advanced similarity search for this article.