A Novel Approach to Urban Village Extraction and Generalization from Digital Line Graphics Using the Computational Geometric Method and the Modified Hausdorff Distance

Xiaorong Gao, Haowen Yan, Xiaomin Lu, Xiaolong Wang, Rong WangISPRS Int. J. Geo Inf 2024 · DOI

TL;DR

This paper introduces a novel method for automatically extracting and generalizing "urban villages" – informal residential areas common in rapidly urbanizing China – from large-scale digital line graphics (DLGs). By leveraging computational geometry and a modified Hausdorff distance to capture their unique morphology, the approach aims to improve map representation and support urban planning decisions.

What Problem Does This Paper Address?

Rapid urbanization in China has led to the proliferation of "urban villages," which are informal, unplanned residential areas often enveloped by modern urban development. These areas are crucial for providing affordable housing but possess distinct spatial characteristics that make them challenging to represent accurately on maps, especially during the map generalization process (reducing detail for smaller scales). Existing research often focuses on structured settlements or uses remote sensing, leaving a gap in automatically extracting and generalizing these informal settlements from vector data like DLGs, which are fundamental to national spatial data infrastructure. The paper aims to bridge this gap by developing a method that can identify and generalize urban villages based on their inherent geometric and spatial properties within vector maps.

Methods

The proposed methodology involves a two-stage process: extraction and generalization.

Extraction: 1. Block Division: The study area is first divided into "blocks" using arterial road networks. This segmentation helps to isolate groups of settlement polygons for analysis. 2. Area Ratio (AreaRatio): For each settlement polygon within a block, an AreaRatio is calculated. This metric compares the area of the settlement polygon to the area of its corresponding Thiessen polygon. A higher AreaRatio suggests a larger proportion of built-up area relative to open space, a characteristic of urban villages. 3. Modified Hausdorff Distance (MHDistance): This is the core novel component. It quantifies the similarity between "facing line segments" (FLSs) of adjacent settlement polygons. FLSs are derived from constrained Delaunay triangulations. The modified Hausdorff distance, adapted from existing work, averages distances between points rather than taking the maximum, making it more robust to local shape variations. This metric is particularly suited to capturing the "closely neighboring" and "similar shape" characteristics of urban village buildings. 4. Thresholding: A two-step thresholding process is used. First, polygons with an AreaRatio above a certain threshold are identified as candidate urban villages. Second, the MHDistance between these candidate urban villages and their neighbors is analyzed. If the MHDistance is below a threshold (indicating similar FLSs and thus close proximity/shared characteristics), neighboring non-urban village polygons are incorporated into the urban village set.

Generalization: After extraction, a generalization method is applied to the identified urban villages and non-urban villages when reducing map scale. This method aims to preserve the semantic distinctiveness of urban villages and built-up districts, ensuring legibility and accurate representation of functional regions, even when road network connectivity is secondary.

Datasets

The study utilizes 1:5K scale Digital Line Graphics (DLGs) data from the Gansu Province Geomatics Center in China. The experimental area is a specific neighborhood in Lanzhou City, selected because it contains both urban village and non-urban village residential areas. This dataset is described as official data, ensuring quality through a rigorous inspection system. No new benchmark datasets were introduced; rather, existing large-scale vector data was used.

Key Insights and Ideas

  • Urban Villages as a Distinct Morphological Class: The paper highlights that urban villages, despite being informal, exhibit consistent morphological traits (high density, small roof areas, narrow roads, close proximity) that can be quantitatively captured from vector data.
  • Leveraging Computational Geometry for Informal Settlements: The application of computational geometry (Delaunay triangulation, Thiessen polygons) and distance metrics (MHDistance) to identify and differentiate informal urban structures is a key innovation.
  • Modified Hausdorff Distance for FLS Similarity: The use of MHDistance specifically for comparing facing line segments of adjacent polygons is crucial. It effectively captures the "sameness" of building edges that face each other across narrow streets, a defining feature of urban villages.
  • Two-Tiered Extraction: The combination of AreaRatio (for density/coverage) and MHDistance (for neighbor similarity) provides a robust approach to distinguishing urban villages from other settlement types.
  • Semantic Preservation in Generalization: The generalization strategy prioritizes maintaining the semantic integrity of different urban functional regions (urban villages, built-up districts) over strict road network connectivity, which is vital for conveying accurate information at smaller scales.

Key Results

  • The proposed method successfully extracted urban village residential areas from a 1:5K DLG dataset in Lanzhou City.
  • The AreaRatio threshold of 0.519689 to 1.348790 was found to be effective in identifying candidate urban villages, balancing omission and commission errors.
  • The MHDistance threshold of 4.4m was identified as indicative of narrow roads (<5m) within urban villages.
  • Qualitative results demonstrated that the generalization method preserved the semantic features of urban villages and built-up districts at reduced scales (1:10K and 1:25K), outperforming a general generalization approach.

Limitations and Future Work

  • Extraction Issues at Edges: The paper acknowledges that polygons at the edge of urban villages can sometimes be misclassified as non-urban villages. This is attributed to special cases in facing line segments, particularly when one segment chain approximates a straight line while the other has a significant turn (e.g., a 90-degree bend). This leads to larger MHDistances and incorrect classification.
  • Generalization Prioritization: The current generalization method prioritizes semantic function regions over road network connectivity. Future work could improve this by incorporating road length into the generalization process.
  • Further Research: Addressing the edge classification issues and refining the generalization method to better consider road network characteristics are identified as key areas for future research. The authors also suggest that if the method proves effective on other regional datasets, it could serve as a reference for urban village generalization across China.

Summary updated

BibTeX
@article{gao2024novel,
  author = {Xiaorong Gao and Haowen Yan and Xiaomin Lu and Xiaolong Wang and Rong Wang},
  title = {{A Novel Approach to Urban Village Extraction and Generalization from Digital Line Graphics Using the Computational Geometric Method and the Modified Hausdorff Distance}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2024},
  doi = {10.3390/ijgi13060198},
}
Figure