ORIENTATION OF OBLIQUE AIRBORNE IMAGE SETS – EXPERIENCES FROM THE ISPRS/EUROSDR BENCHMARK ON MULTI-PLATFORM PHOTOGRAMMETRY

M. Gerke, Francesco Nex, Fabio Remondino, Karsten Jacobsen, Jens Kremer, Wilfried Karel, Han Hu, Wojciech Ostrowski

Abstract

During the last decade, the use of airborne multi-camera systems has increased significantly. These systems enable image capture under different angles, providing visibility of lateral parts of elevated objects like buildings or trees. However, varying scale, occlusion, and perspective differences challenge image processing. From an image orientation point of view, these multi-camera systems offer the advantage of better ray intersection geometry compared to nadir-only image blocks. This paper presents results from the ISPRS/EUROSDR benchmark on multi-platform photogrammetry, focusing on tie point matching across viewing directions, the influence of oblique images on bundle adjustment, and the roles of image overlap and Ground Control Point (GCP) distribution. We confirm that height accuracy is significantly improved when all images are used and that high overlap (80/80%) provides much better object space accuracy compared to 60/60% overlap. Furthermore, modern commercial software packages perform exceptionally well on these datasets.

Introduction

Traditional aerial photogrammetry relies on nadir (straight-down) imagery, which is effective for 2D mapping but limited for 3D modeling of vertical structures. Multi-camera systems, such as the IGI Pentacam, address this by capturing one nadir and four oblique views simultaneously. While these systems provide better ray intersection geometry, they introduce challenges like perspective distortion, scale variation, and occlusions. This paper evaluates these factors using the ISPRS/EuroSDR benchmark dataset to determine how they impact the final accuracy of the 3D bundle adjustment.

Methodology and Data

The study utilizes a dataset from Dortmund, Germany, comprising 1260 images captured with an IGI Pentacam. The analysis compares two primary configurations: a "full" set with 80% forward and 80% side overlap, and a "reduced" set with 60%/60% overlap. Ground truth was established using 33 high-precision RTK-GNSS points, with 10 used as GCPs and 23 as independent check points. Several software packages, including commercial SfM suites like Pix4Dmapper and research tools like BLUH, were evaluated for their performance in Bundle Block Adjustment (BBA) with self-calibration.

Results and Analysis

The Advantage of Oblique Geometry

A key finding is that oblique images significantly enhance vertical accuracy. By providing steeper ray intersection angles, the mathematical solution for the Z-axis becomes more stable. Experimental results confirm that vertical accuracy is significantly better when the full image set is used. As shown in the comparison of RMSE values, the Z-accuracy for the nadir-only configuration is nearly three times worse than that of the full Pentacam set at ground control points.

RMSE comparison between nadir-only and full Pentacam image sets.

Impact of Image Overlap and GCP Distribution

Image overlap serves as a critical buffer for accuracy. High overlap (80/80%) provides enough redundant data to mask systematic errors or suboptimal GCP placement. In contrast, 60/60% overlap configurations are much more sensitive to GCP distribution. Moving from 60/60% to 80/80% overlap can reduce random errors by a factor of 2 to 3, bringing the Z-RMSE down to approximately 10cm (roughly 1 GSD). At lower overlaps, poor GCP distribution can cause the Z-error to spike significantly, jumping from 16cm to 25cm in certain scenarios.

Impact of overlap and GCP scenarios on vertical accuracy (RMSE in Z).

Discussion and Conclusion

The study highlights that while matching across opposing viewing directions remains an algorithmic challenge, current SfM software is highly capable of handling structured aerial blocks. Commercial packages often match or exceed the precision of specialized research prototypes. The results underscore that high overlap and the inclusion of oblique imagery are essential for achieving high-precision vertical accuracy. Future work will focus on improving automated matching across diverse perspectives and refining physical models for multi-camera rig synchronization.

Summary updated

Abstract

During the last decade the use of airborne multi camera systems increased significantly. The development in digital camera technology allows mounting several mid- or small-format cameras efficiently onto one platform and thus enables image capture under different angles. Those oblique images turn out to be interesting for a number of applications since lateral parts of elevated objects, like buildings or trees, are visible. However, occlusion or illumination differences might challenge image processing. From an image orientation point of view those multi-camera systems bring the advantage of a better ray intersection geometry compared to nadir-only image blocks. On the other hand, varying scale, occlusion and atmospheric influences which are difficult to model impose problems to the image matching and bundle adjustment tasks. In order to understand current limitations of image orientation approaches and the influence of different parameters such as image overlap or GCP distribution, a commonly available dataset was released. The originally captured data comprises of a state-of-the-art image block with very high overlap, but in the first stage of the so-called ISPRS/EUROSDR benchmark on multi-platform photogrammetry only a reduced set of images was released. In this paper some first results obtained with this dataset are presented. They refer to different aspects like tie point matching across the viewing directions, influence of the oblique images onto the bundle adjustment, the role of image overlap and GCP distribution. As far as the tie point matching is concerned we observed that matching of overlapping images pointing to the same cardinal direction, or between nadir and oblique views in general is quite successful. Due to the quite different perspective between images of different viewing directions the standard tie point matching, for instance based on interest points does not work well. How to address occlusion and ambiguities due to different views onto objects is clearly a non-solved research problem so far. In our experiments we also confirm that the obtainable height accuracy is better when all images are used in bundle block adjustment. This was also shown in other research before and is confirmed here. Not surprisingly, the large overlap of 80/80% provides much better object space accuracy – random errors seem to be about 2-3fold smaller compared to the 60/60% overlap. A comparison of different software approaches shows that newly emerged commercial packages, initially intended to work with small frame image blocks, do perform very well.
BibTeX
@inproceedings{gerke2016orientation,
  author = {M. Gerke and Francesco Nex and Fabio Remondino and Karsten Jacobsen and Jens Kremer and Wilfried Karel and Han Hu and Wojciech Ostrowski},
  title = {{ORIENTATION OF OBLIQUE AIRBORNE IMAGE SETS – EXPERIENCES FROM THE ISPRS/EUROSDR BENCHMARK ON MULTI-PLATFORM PHOTOGRAMMETRY}},
  booktitle = {˜The},
  year = {2016},
  doi = {10.5194/isprs-archives-xli-b1-185-2016},
}
Figure