A Reconstruction Method for Broken Contour Lines Based on Similar Contours

Chengming Li, Xiaoli Liu, Wei Wu, Zhiwei HaoISPRS Int. J. Geo Inf 2019 · DOI

TL;DR

This paper introduces a novel method for reconstructing broken contour lines in topographic maps by leveraging similar, complete contour lines as references. By employing the discrete Fréchet distance for similarity measurement and a node densification technique, the method significantly improves the accuracy and robustness of contour line reconstruction, especially in complex terrains and severe break situations, outperforming traditional approaches.

What Problem Does This Paper Address?

Topographic maps are crucial for creating digital elevation models (DEMs) and 3D simulations. Contour lines are a fundamental element of these maps. However, during the extraction of contour lines from scanned maps, gaps or breaks frequently occur due to overlays or intersections with other map features. These breaks pose a significant challenge for automated DEM generation and 3D modeling. Traditional methods for reconstructing these broken contour lines, often based on simple geometric properties like Euclidean distance and direction, struggle when contour lines are severely broken or the underlying terrain is complex (e.g., ridges, valleys, saddles). In such scenarios, these methods are prone to incorrect connections and missed segments, necessitating extensive manual correction and reducing their practical utility.

Methods

The proposed method addresses the limitations of existing continuity-based reconstruction techniques by introducing a "reference line" approach. The core methodology involves three main steps:

  1. Contour Line Node Densification: To improve the accuracy of similarity measurement, especially when contour nodes are sparse, this step uniformly inserts new points along the contour line. This ensures that the discrete Fréchet distance calculation more closely approximates the continuous Fréchet distance, leading to a more accurate assessment of similarity.
  2. Reference Line Selection using Fréchet Distance: The method identifies a complete contour line that is spatially close to and has a similar trend as the broken contour line. This is achieved by calculating the discrete Fréchet distance between segments of the broken contour line and surrounding complete contour lines. The contour line with the minimum Fréchet distance is selected as the reference line. This approach is more robust than simple Euclidean distance, as it considers the overall shape similarity between curves.
  3. Interpolation and Connection Based on the Reference Line: Once a reference line is established, the broken contour line is interpolated and connected. This involves determining the projected position of the broken contour line onto the reference line, identifying the extended direction, and using elevation information to calculate interpolated points. These points are then connected sequentially to form a complete contour line. The interpolation uses a proportion-based approach derived from the elevation differences between the broken contour and its reference lines.

Datasets

The method was validated using actual contour line data from a mountainous and hilly area in Yunnan Province, China. This dataset covered an area of 29.1 km² and contained 5948 contour lines with a 5m interval, of which 3277 were broken contour lines (55% of the total). No new benchmark datasets were introduced; the study utilized existing real-world topographic data.

Key Insights and Ideas

  • Fréchet Distance for Shape Similarity: Moving beyond simple point-to-point distances, the discrete Fréchet distance provides a more robust measure of curve similarity, capturing the overall shape and alignment of contour lines. This is crucial for identifying appropriate reference lines.
  • Reference Line as a Geometric Anchor: Using a complete contour line as a reference acts as a geometric anchor, guiding the reconstruction process and providing context that is lost in methods relying solely on local breakpoint information.
  • Node Densification for Accuracy: The insight that sparse nodes can lead to inaccurate similarity measurements highlights the importance of pre-processing contour lines to ensure sufficient detail for accurate geometric comparisons.
  • Improved Robustness in Complex Scenarios: The method demonstrates a significant improvement in handling severely broken contour lines and complex terrains, areas where traditional methods falter due to their limited consideration of geometric context.

Key Results

The experimental results demonstrated the effectiveness of the proposed method:

  • High Reconstruction Accuracy: The overall reconstruction accuracy of broken contour lines was 93.5%.
  • Improved Performance in Complex Cases: The method showed significantly improved reconstruction accuracy in complex broken contour zones compared to the minimum point pair method.
  • Universality: The method proved applicable to both general and complex broken contour lines, indicating good universality.
  • Smoother and Non-Intersecting Lines: Compared to the minimum point pair method, the reconstructed contour lines were more uniform in distance and smoother, with fewer instances of self-intersection.

Limitations and Future Work

The authors acknowledge a potential limitation:

  • Leakage Connection: A "leakage connection" (presumably meaning an incorrect or unintended connection) might occur if the similarity between the broken contour and the selected reference contour is low.

Future research directions suggested by the authors include:

  • Reference Contour Selection and Optimization: Further work will focus on improving the selection and optimization of reference contours to enhance reconstruction accuracy.

Summary updated

BibTeX
@article{li2019reconstruction,
  author = {Chengming Li and Xiaoli Liu and Wei Wu and Zhiwei Hao},
  title = {{A Reconstruction Method for Broken Contour Lines Based on Similar Contours}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2019},
  doi = {10.3390/ijgi8010008},
}
Figure