OBLIQUE MULTI-CAMERA SYSTEMS – ORIENTATION AND DENSE MATCHING ISSUES

Ewelina Rupnik, Francesco Nex, Fabio Remondino

Abstract

The use of oblique imagery has become a standard for many civil and mapping applications, thanks to the development of airborne digital multi-camera systems. The indisputable virtue of oblique photography lies in its simplicity of interpretation and understanding for inexperienced users, allowing their use in applications such as building reconstruction, structural damage classification, and land updating. This paper reports an overview of actual oblique commercial systems and presents a workflow for the automated orientation and dense matching of large image blocks. Perspectives, potentialities, pitfalls, and suggestions for achieving satisfactory results are given, alongside tests performed on urban datasets.

Introduction

For decades, aerial photogrammetry relied on "nadir" (vertical) imagery as the industry standard. However, nadir views are inherently limited, as they cannot capture building facades or the "bird's eye" perspectives required for modern 3D city modeling and disaster assessment. The transition to oblique imagery—where cameras are tilted at 40-50°—introduces significant technical challenges. Traditional photogrammetric software often struggles with the extreme scale changes, severe occlusions, and complex image footprints produced by multi-camera rigs. Without specialized workflows, mathematical models often fail to converge, resulting in distorted maps or incomplete 3D models. This research addresses the need for an automated, reliable pipeline to process these complex datasets.

Methodology

The proposed workflow utilizes a combination of commercial and open-source tools, specifically Apero for orientation and MicMac for dense matching. To overcome the geometric complexities of oblique data, several key technical strategies are implemented:

  • Image Connectivity Graphs: To manage thousands of images efficiently, the system avoids exhaustive "all-to-all" matching. Instead, GNSS/IMU data is used to construct a connectivity graph. Images are only linked if they share a ground footprint and have similar "look directions," ensuring stability during bundle adjustment.
  • Directional Master Selection: For dense 3D reconstruction, images are grouped by their cardinal direction (North, South, East, West). "Master" images are selected for each direction to ensure comprehensive reconstruction of all building sides before the final results are merged.
  • Mixed Pixel Filtering: Oblique perspectives often generate noise near building edges where camera rays graze surfaces. A geometric filter is applied to remove points where the local surface normal is nearly parallel to the camera's line of sight.
Automated orientation and camera network connectivity graph for an urban block.

Experimental Setup

The study evaluates two primary hardware configurations using urban datasets:

  1. Maltese-cross System (Midas): A rig featuring one vertical camera and four tilted cameras oriented toward cardinal directions.
  2. Fan System (VisionMap A3): A "stepping frame" system that sweeps across the flight path, creating a massive 109° field of view. The evaluation focused on a 0.5 x 0.5 km urban block to test connectivity graph logic and dense point cloud generation.

Discussion

The research highlights several critical insights into oblique imaging:

  • System Configuration: The "Maltese-cross" configuration is currently preferred in the industry due to its modularity, allowing for a flexible mix of sensor formats.
  • Stability through Selection: Smart image selection via connectivity graphs is not merely a speed optimization but a requirement for mathematical stability. Without it, bundle adjustments are prone to divergence.
  • Data Quality vs. Quantity: While oblique systems produce point clouds 2–3 times denser than nadir flights, the tilted geometry increases noise. Effective filtering that accounts for camera perspective relative to surface geometry is essential for high-quality outputs.

Results

The automated workflow successfully oriented large blocks of oblique images (e.g., 280 images from a VisionMap A3) that typically fail in standard software. A primary achievement was the successful reconstruction of vertical building facades, which are missed by nadir-only systems. The workflow produced high-density point clouds (theoretically up to 100 points/m²) with a Ground Sample Distance (GSD) of approximately 10cm, meeting the requirements for detailed 3D city modeling.

Dense 3D reconstruction showing detailed building facades from oblique imagery.

Conclusion

Despite the successes, several limitations remain. Large glass facades and repetitive patterns continue to cause noise and "holes" in 3D models. Furthermore, the computational burden of processing billions of points remains a bottleneck for standard hardware. Future work will focus on refining the merging process to eliminate radiometric inconsistencies and artifacts. Additionally, the industry currently lacks formal benchmarks for comparing oblique systems, a gap the authors are addressing through international collaborations with EuroSDR and ISPRS.

Summary updated

Abstract

The use of oblique imagery has become a standard for many civil and mapping applications, thanks to the development of airborne digital multi-camera systems, as proposed by many companies (Blomoblique, IGI, Leica, Midas, Pictometry, Vexcel/Microsoft, VisionMap, etc.). The indisputable virtue of oblique photography lies in its simplicity of interpretation and understanding for inexperienced users allowing their use of oblique images in very different applications, such as building detection and reconstruction, building structural damage classification, road land updating and administration services, etc. The paper reports an overview of the actual oblique commercial systems and presents a workflow for the automated orientation and dense matching of large image blocks. Perspectives, potentialities, pitfalls and suggestions for achieving satisfactory results are given. Tests performed on two datasets acquired with two multi-camera systems over urban areas are also reported.
BibTeX
@inproceedings{rupnik2014oblique,
  author = {Ewelina Rupnik and Francesco Nex and Fabio Remondino},
  title = {{OBLIQUE MULTI-CAMERA SYSTEMS – ORIENTATION AND DENSE MATCHING ISSUES}},
  booktitle = {˜The},
  year = {2014},
  doi = {10.5194/isprsarchives-xl-3-w1-107-2014},
}
Figure