LandParser: Intelligent Land Boundary Detection and Encroachment Analysis Using Deep Learning with Satellite Imagery

Abstract

Land management is a critical concern in modern governance, particularly with the increase in urbanisation and disputes over land ownership. This initiative introduces a comprehensive intelligent system aimed at achieving three main goals: detecting land encroachment, parsing land and recognising boundaries, and predicting land prices. The system utilises satellite images and user-submitted photos, employing deep learning methods, especially U-Net-based semantic segmentation, to precisely identify and mark encroached areas. For land parsing, it uses geospatial data and image processing techniques to accurately define land boundaries and improve the precision of digital land records. Furthermore, machine learning models like XGBoost are used to forecast land prices by considering various factors such as location, nearby infrastructure, land use category, and historical trends. The system is designed with an interactive modular architecture that integrates GIS tools, OpenCV, and real-time analytics to facilitate efficient visualisation and decision-making. By combining computer vision, geospatial intelligence, and predictive modelling, the platform provides a scalable solution for government officials, urban planners, and real estate stakeholders.

Introduction

LandParser is an innovative geo-intelligent system designed for automated land boundary detection, encroachment analysis, and land valuation using deep learning and satellite/UAV imagery. It addresses critical inefficiencies in Maharashtra’s current land management practices, which rely heavily on manual surveys and fragmented data, causing delays and errors in detecting illegal encroachments and maintaining accurate land records.

Key Components and Methodology:

• Land Boundary Detection:
  Utilizes the advanced U-Net3++ deep CNN architecture for precise semantic segmentation of cadastral parcel boundaries from high-resolution satellite and UAV imagery, achieving high accuracy in delineating complex boundaries.

• Encroachment Detection:
  Employs temporal change detection by comparing segmented parcel maps over time to flag unauthorized land-use changes, enhancing monitoring and enforcement efficiency.

• Land Valuation:
  Uses machine learning ensemble models, especially XGBoost, to predict land prices based on spatial, agronomic, infrastructural, and market features, improving valuation accuracy compared to traditional methods.

• Ownership Tagging:
  Integrates government records with satellite data to map and update land ownership information in real time.

Results and Validation:

• Boundary detection achieved high segmentation accuracy (IoU ~91.4%, pixel accuracy ~96.8%).

• Encroachment detection showed 92% precision and 88% recall, with false positives mostly from seasonal land-use changes.

• Land price prediction had an R² score of 0.89 with low error rates, benefiting from rich feature integration.

• The system includes an interactive GIS dashboard for visualizing boundaries, encroachments, ownership, and valuation data, receiving positive user feedback for usability and decision support.

Contributions and Significance:

• Combines CNN-based segmentation, temporal change analysis, and machine learning valuation into a unified, user-friendly platform.

• Improves speed, accuracy, and transparency in land governance and encroachment monitoring.

• Demonstrates how integrating multi-source data and AI advances traditional cadastral management systems.

Conclusion

LandParser represents a significant advancement in land management, strategically leveraging advanced AI/ML and remote sensing technologies to automate critical tasks that have historically relied on inefficient manual methods. Its capacity to provide real-time insights into land boundaries, encroachments, and valuations, complemented by an intuitive interactive dashboard, promises enhanced transparency, improved precision, and truly data-driven decision-making for urban planning and governance. The project, spearheaded by the Maharashtra Remote Sensing Application Centre (MRSAC), is well-aligned with national digital land administration initiatives, such as the Digital India Land Records Modernisation Programme (DILRMP), and mirrors global trends in the adoption of geospatial AI by governmental bodies.

References

[1] A. Alshari, M. B. Abdulkareem, and B. W. Gawali, \"Classification of land use/land cover using artificial intelligence (ANN-RF),\" Front. Artif. Intell., vol. 5, pp. 1–14, Jan. 2023, doi: 10.3389/frai 2022.964279 [2] M. C. Taberner, F. J. García-Haro, B. Martínez, E. Izquierdo-Verdiguier, C. Atzberger, G. Camps-Valls, and M. A. Gilabert, \"Understanding deep learning in land use classification based on Sentinel-2 time series,\" Sci. Rep., vol. 10, no. 1, pp. 1–12, Nov. 2020, doi: 10.1038/s41598-020-74215-5. [3] J. Kim, J. Won, H. Kim, and J. Heo, \"Machine-learning-based prediction of land prices in Seoul, South Korea,\" Sustainability, vol. 13, no. 23, pp. 1–18, Nov. 2021, doi: 10.3390/su132313088. [4] B. Fetai, M. Racic, and A. Lisec, \"Deep learning for detection of visible land boundaries from UAV imagery,\" Remote Sens., vol.. 13, no. 11, pp. 1–17, May 2021, doi: 10.3390/rs13112077. [5] S. Crommelinck, M. Koeva, M. Y. Yang, and G. Vosselman, \"Application of deep learning for delineation of visible cadastral boundaries from remote sensing imagery,\" Remote Sens., vol. 11, no. 21, pp. 1–22, Nov. 2019, Doi: 10.3390/rs11212505. [6] T. Lu, L. Wan, S. Qi, and M. Gao, \"Land cover classification of UAV remote sensing based on Transformer–CNN hybrid architecture,\" Sensors, vol. 23, no. 11, p. 5288, Jun. 2023, doi: 10.3390/rs23115288. [7] A. Vali, S. Comai, and M. Matteucci, \"Deep learning for land use and land cover classification based on hyperspectral and multispectral earth observation data: A review,\" Remote Sens., vol. 12, no. 15, p. 2495, Jul. 2020, doi: 10.3390/rs12152495. [8] X.-Y. Tong, G.-S. Xia, Q. Lu, H. Shen, S. Li, S. You, and L. Zhang, \"Land-cover classification with high-resolution remote sensing images using transferable deep models,\" Remote Sensing of Environment (RSE), vol.. 237 Article 111322, Feb. 2020, doi: 10.1016/j.rse.2019.111322. [9] M. Brimble, T. Haithcoat, and G. J. Scott, \"Using machine learning and remote sensing to value property in Kigali, Rwanda,\" Working Paper, Feb. 2020. [10] F. Tanrikulu, T. Haithcoat, and G. J. Scott, \"Land valuation using an innovative model combining machine learning and spatial context,\" M.S. thesis, Univ. of Missouri, Columbia, MO, USA, 2022, doi: 10.1109/IGARSS53475.2024.10642699. [11] \"Challenges and Limitations of Remote Sensing Applications in Northern Peatlands: Present and Future Prospects,\" MDPI, accessed August 12, 2025. [Online]. [12] \"Deep Learning for Detection of Visible Land Boundaries from UAV,\" accessed August 12, 2025. [13] \"Application of Deep Learning for Delineation of Visible Cadastral,\" accessed August 12, 2025. [14] \"State-of-the-Art Deep Learning Methods for Object Detection in Remote Sensing Satellite Images,\" PMC, accessed August 12, 2025. [15] \"Deep Learning-Based Semantic Segmentation of Urban Features in Satellite Images: A Review and Meta-Analysis,\" MDPI, accessed August 12, 2025. [Online] [16] \"satellite-image-deep-learning/techniques,\" GitHub, accessed August 12, 2025. [Online]. [17] \"Detecting Cadastral Boundary from Satellite Images Using U-Net model,\" arXiv, accessed August 12, 2025. [Online]. [18] \"Detecting Cadastral Boundary from Satellite Images Using U-Net model,\" ResearchGate, accessed August 12, 2025. [Online]. [19] \"Encroachment Detection,\" Satpalda, accessed August 12, 2025. [Online]. [20] \"Satellite Imagery and Deep Learning for Smart City Insights,\" Medium, accessed August 12, 2025.

Copyright

Copyright © 2025 Dr. Reema Roychaudhary, Manasvi Giradkar, Radhika Salodkar, Gargi Gundawar, Gargi Udapure. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.