A METHOD FOR THE SEAMLINES NETWORK AUTOMATIC SELECTION BASED ON BUILDING VECTOR

Puda Li, Yanwen Dong, Ye Hu, X. Li, Pang‐Ning Tan2018 · DOI

Abstract

In order to improve the efficiency of large-scale urban orthophoto production, this paper presents a method for the automatic selection of seamline networks based on building vector data. By combining building vectors, heights, and Digital Elevation Models (DEM), a simple 3D model is constructed to determine the imaging area of structures on single Digital Orthophoto Maps (DOM). An initial Voronoi network is generated based on image centers and subsequently optimized at both node and seamline levels to bypass building imaging areas. Experimental results indicate that this method effectively avoids building intersections while maintaining the geometric advantages of Voronoi-based mosaicking, offering a robust solution for automated image stitching in urban environments.

Introduction

Creating high-quality Digital Orthophoto Maps (DOM) requires "mosaicking" numerous aerial images. A central challenge in this process is the selection of seamlines—the boundaries where adjacent images are joined. In urban areas, tall buildings are subject to relief displacement, causing them to appear to "lean" away from the image center. If a seamline intersects such a building, the resulting mosaic often suffers from "ghosting" or structural misalignments. While manual seamline selection can mitigate these issues, it is prohibitively slow for large-scale city mapping. This paper proposes an automated workflow to intelligently reroute seamlines around buildings using existing GIS vector data.

Methodology

The proposed method follows a structured pipeline to transform raw aerial imagery and vector data into an optimized seamline network.

Workflow of the proposed seamline selection method.

The process consists of four primary stages:

  1. Digital Building Model (DBM) Construction: Building footprints from Digital Line Graphs (DLG) are combined with height data and DEMs to create simplified 3D representations of urban structures.
  2. Imaging Area Calculation: Using collinearity equations, the system determines the 2D projection (imaging area) of each 3D building on the orthophoto. This identifies regions where seamlines should not pass.
  3. Initial Voronoi Network Generation: A standard Voronoi diagram is created based on image centers. This ensures that each pixel in the final mosaic is sourced from the image where it was captured with the least perspective distortion.
  4. Two-Stage Optimization:
    • Node Optimization: Vertices in the Voronoi network that fall within building imaging areas are shifted to the nearest ground-level position.
  5. Seamline Optimization: Edges that cross building imaging areas are rerouted to follow building boundaries, ensuring no structure is "sliced" by a seam.

Key Insights: The "Leaning" Building Problem

A critical insight of this research is the distinction between a building's ground footprint and its "imaging area." Because of perspective projection, a building occupies more space in an aerial photo than its base alone.

Projection of a building from 3D space to the 2D imaging area on a DOM.

By calculating this extended imaging area, the algorithm can proactively avoid the entire visual extent of the building, preventing the "broken building" artifacts common in simpler automated mosaicking techniques. Furthermore, by starting with a Voronoi network and only applying local "nudges," the method preserves the global geometric integrity of the mosaic without the high computational cost of pure pathfinding algorithms.

Results and Discussion

The method was validated using 32 UAV images (0.15m resolution) covering an area with 526 buildings. The automated system successfully generated a seamline network that "snaked" through streets and alleys, completely avoiding building rooftops.

While highly effective, the method's performance is tied to the accuracy of the input GIS building vector data. Missing or outdated vector information can lead to seamline errors. Future work may involve integrating deep-learning-based building detection to reduce reliance on pre-existing databases and handle complex architectural features more accurately.

Conclusion

The proposed building-vector-based approach significantly enhances the automation and quality of urban orthophoto production. By intelligently navigating the "imaging areas" of structures, it provides a scalable solution for creating seamless, distortion-minimum digital maps of complex cityscapes.

Summary updated

Abstract

In order to improve the efficiency of large scale orthophoto production of city, this paper presents a method for automatic selection of seamlines network in large scale orthophoto based on the buildings’ vector. Firstly, a simple model of the building is built by combining building’s vector, height and DEM, and the imaging area of the building on single DOM is obtained. Then, the initial Voronoi network of the measurement area is automatically generated based on the positions of the bottom of all images. Finally, the final seamlines network is obtained by optimizing all nodes and seamlines in the network automatically based on the imaging areas of the buildings. The experimental results show that the proposed method can not only get around the building seamlines network quickly, but also remain the Voronoi network’ characteristics of projection distortion minimum theory, which can solve the problem of automatic selection of orthophoto seamlines network in image mosaicking effectively.
BibTeX
@inproceedings{li2018method,
  author = {Puda Li and Yanwen Dong and Ye Hu and X. Li and Pang‐Ning Tan},
  title = {{A METHOD FOR THE SEAMLINES NETWORK AUTOMATIC SELECTION BASED ON BUILDING VECTOR}},
  booktitle = {˜The},
  year = {2018},
  doi = {10.5194/isprs-archives-xlii-3-909-2018},
}
Figure