Choosing the right nadir image overlap for optimal DSM and True Ortho generation
The image overlap ratio significantly impacts the quality of reality products. Optimal image overlap can minimize artifacts, such as data gaps and improve both accuracy and sharpness. Additionally, it is closely linked to the performance, significantly influencing processing runtime.
In certain scenarios, increasing the overlap ratio results in substantial improvements, while in others, the benefits are less pronounced and may lead to increased processing efforts. Therefore, it is essential to find a balance between quality and efficiency. This article provides recommendations for selecting the right image overlap to achieve this balance. As the flight planning for oblique projects should align with its nadir camera, these recommendations can also be applied to oblique setups.
Understanding flight patterns and image overlaps
Nadir images are typically acquired from flights following a regular flight pattern, such as parallel strips. These images are characterized by two key parameters: a forward (in-strip) overlap and a sideward (cross-strip) overlap of adjacent images. To ensure high coverage and minimize processing time, images on adjacent strips should be arranged in a rectilinear distribution. A chessboard pattern is recommended to achieve this.
In addition to the flight pattern, the camera's opening angle needs to be taken into account when determining the overlap parameter for a photo flight. It affects the extent of occlusions occurring towards the image borders. This article focuses on standard large-format cameras, which typically have an opening angle of 60-70 degrees across flight direction.
Selecting the right image overlap for a scene
The optimal image overlap is highly dependent on the characteristics of the scene being captured. Urban areas, characterized by densely packed, high-rise buildings require a higher image overlap to compensate for occlusions. In contrast, rural regions, which are often sparsely built, can be adequately covered with less redundancy.
In the following sections, you will find suggested configurations for different scenes.
Minimum configuration: 60% forward overlap / 30% sideward overlap
Historically, nadir images were acquired with a 60/30 overlap to facilitate high-quality aerotriangulation while minimizing the costs associated with analog image capturing, such as film and processing expenses.
Using an overlap of 60/30, a redundancy of at most two images is possible for the central image area.
While this configuration is suitable for traditional datasets, it is generally advisable to utilize a forward overlap of at least 80% for modern dense matching applications.
Recommended minimum configuration: 80% forward overlap / 30% sideward overlap
Increasing the forward overlap to 80% significantly reduces occlusions and enhances redundancy without necessitating additional flight efforts. Each inner image then has four suitable stereo partners. The substantial reduction in stereo occlusions – areas where pixels are not being observed in multiple stereo models – leads to fewer data gaps. Additionally, the increased image similarity associated with the 80% forward overlap model enhances matching performance, resulting in improved edge recovery.
This configuration of 80/30 is optimal for capturing open landscapes from higher altitudes. However, in scenarios with more complex topography, such as urban areas with buildings and increased occlusions, it is advisable to also increase the sideward overlap. This adjustment helps to resolve stereo occlusions between strips, especially in cases where streets run perpendicular to the flight direction.
In the 80% in-strip, 30% cross-strip overlap configuration the central image area (red) is covered by at least 5 images (4 stereo models). The redundancy enables outlier filtering, noise and occlusion reduction.In the 80% in-strip, 30% cross-strip overlap configuration the central image area (red) is covered by at least 5 images (4 stereo models). The redundancy enables outlier filtering, noise and occlusion reduction.
Urban scenes without high-rise buildings: 80% forward overlap / 60% sideward overlap
Using an 80/60 overlap configuration effectively reduces stereo occlusions between strips, which are typically caused by the relief displacement effect of buildings. With a 60% side overlap, an additional stereo model is incorporated between strips, resulting in four models in-strip and two cross-strip. This setup enables the generation of more points in areas prone to stereo occlusions, such as ground-level surfaces adjacent to buildings. Consequently, this overlap configuration significantly improves the quality of True Orthophotos, as accurately captured points in front of building facades are crucial for achieving well-defined edges.
Urban scenes with high-rise buildings: 80% forward overlap / 80 % sideward overlap
An 80/80 overlap configuration is recommended to address the strong stereo occlusions caused by tall buildings. By increasing both the forward and side overlaps, this configuration ensures better coverage and improved data quality in densely built environments.
Generic approach: Central Image Contribution
A more generalized method, independent of the camera type, is defined by considering the Central Image Contribution. The concept focuses on the central region of the image, which is determined by the area between overlaps. The image overlap is set in relation to the building's lean to minimize occlusions caused by the building’s relief displacement. The effective maximum occlusion can be estimated using the opening angle between the edges of the Central Image Contribution and a maximum building height.
Summary
The overlap of images has a significant influence on the resulting products. When planning a capture session, image overlaps should be chosen considering the characteristics of the area of interest. By this, high-quality reconstructions can be achieved even for dense, high-rising areas, while processing efforts are kept at a minimum for scenes with few occlusions.
The following table provides an overview of recommended overlap configurations for various urban scenarios.
In-strip/cross-strip Overlap (%) | Key Attributes | Recommended Scene Type |
80/80 | Greatly reduces cross-strip stereo occlusions | Urban scenes with high-rises |
80/60 | Reduces cross-strip stereo occlusions | Urban scenes without high-rises |
80/30 | Recommended minimum coverage. Four suitable stereo pairs per image | Rural scenes from high altitudes |
60/30 | Minimum coverage required. The central image area is only covered by two images | Rural scenes from high altitudes |