Abstract Nowadays, multi-camera platforms combining nadir and oblique cameras are experiencing a revival. Due to their advantages such as ease of interpretation, completeness through mitigation of occluding areas, as well as system accessibility, they have found their place in numerous civil applications. However, automatic post-processing of such imagery still remains a topic of research. Configuration of cameras poses a challenge on the traditional photogrammetric pipeline used in commercial software and manual measurements are inevitable. For large image blocks it is certainly an impediment. Within the theoretical part of the work, we review three common least square adjustment methods and recap on possible ways for a multi-camera system orientation. In the practical part, we present an approach that successfully oriented a block of 550 images acquired with an imaging system composed of 5 cameras (Canon Eos 1D Mark III) with different focal lengths. Oblique cameras are rotated in the four looking directions (forward, backward, left and right) by 45° with respect to the nadir camera. The workflow relies only upon open-source software: a developed tool to analyse image connectivity and Apero to orient the image block. The benefits of the connectivity tool are twofold: in terms of computational time and success of Bundle Block Adjustment. It exploits the georeferenced information provided by the Applanix system in constraining feature point extraction to relevant images only, and guides the concatenation of images during the relative orientation. Ultimately an absolute transformation is performed resulting in mean re-projection residuals equal to 0.6 pix.
Introduction
Oblique aerial imagery—images captured at an angle rather than straight down—is undergoing a significant revival in urban mapping. Unlike traditional nadir (top-down) photography, oblique views capture building facades and narrow urban canyons, providing a more complete 3D representation of the environment. However, processing these images automatically remains a significant hurdle in the photogrammetric pipeline.
Standard software often struggles with the geometric complexity of oblique images, which exhibit massive scale variations and perspective distortions within a single frame. Furthermore, the computational scale of modern datasets is immense; large blocks of images require finding matching "tie points" across thousands of pairs. Without an intelligent method to predict image overlap, software wastes significant time attempting to match images that do not share common features. Consequently, many commercial solutions still require human operators to manually seed the orientation process.
Methodology and the Connectivity Tool
This paper proposes a fully automatic workflow utilizing the open-source MicMac/Apero ecosystem, enhanced by a custom-developed connectivity tool. The primary goal of this tool is to prune the search space for feature extraction and guide the mathematical orientation process.
Footprint Projection and Spatial Indexing
The workflow begins by using rough GNSS/IMU data from the aircraft to project the "footprint" of every image onto a simplified ground model. To manage the spatial data efficiently, a Kd-Tree search is employed to quickly identify neighboring camera stations within a specific radius. This prevents the system from performing an exhaustive $N^2$ comparison of all images.
Smart Filtering and Guided Concatenation
Images are only matched if they exhibit significant overlap and share compatible viewing angles. For instance, a forward-looking oblique camera is matched with other forward-looking cameras or the nadir camera, but not with backward-looking ones. This logic is formalized into a "concatenation queue" that dictates the order in which images are linked during the Bundle Block Adjustment (BBA).
By treating nadir images as the primary "anchors," the workflow ensures geometric consistency. Oblique images are tied to these anchors or to other obliques with similar perspectives, maintaining a stable mathematical model as the block grows.
Experimental Results
The methodology was validated using a dataset captured over Milano, Italy. The hardware consisted of a Midas-BlomOblique system with five Canon EOS 1D Mark III cameras: one 80mm nadir lens and four 100mm oblique lenses tilted at 45°. The block comprised 550 images covering an 8 km x 3.5 km area with high redundancy (80% forward and 50% side overlap).
Efficiency and Accuracy
The implementation of the connectivity tool resulted in a dramatic reduction in processing time. By constraining feature point extraction to relevant image pairs, the authors achieved a speedup of 15 to 20 times compared to unconstrained matching.
In terms of precision, the final BBA achieved a mean re-projection residual of 0.5 pixels for relative orientation and 0.6 pixels after absolute transformation. The system successfully linked all 550 images into a single coherent block without the need for manual Ground Control Points (GCPs), relying entirely on the onboard Applanix GNSS/IMU system for global positioning.
Discussion and Conclusion
The results demonstrate that open-source tools like Apero, when paired with intelligent pre-processing logic, can handle industrial-scale photogrammetric tasks that were previously the domain of expensive proprietary software.
However, some limitations remain. The current 0.6-pixel accuracy represents internal consistency (residuals) rather than absolute geographic accuracy against ground truth. Future work will involve validation against external GCPs. Additionally, the authors suggest that incorporating "lever arm" constraints—the fixed physical rotations and distances between cameras on the same rig—could further stabilize the block and reduce mathematical unknowns. Testing the tool across more diverse topographies will also be essential to prove its general versatility.
Summary updated