Production with SURE - Best Practices
Doing professional production with SURE is simple.
Following these simple guidelines enables you to get the best quality results in the shortest time possible.
Use hardware that follows our recommendations
Use oriented (aero-triangulated) imagery. SURE supports a wide range of orientation formats (Inpho prj, gori, ...). The production of high-quality results requires high-quality orientations (error below 2 pixels). Direct georeferencing can usually not meet this requirement.
To acquire the best possible geometric quality out of Aerial Datasets, process them using the quality level Ultra. The depth accuracy of the reconstructed points decreases by a factor of two each time the quality is reduced by one level. A lower depth accuracy leads to a decrease in edge sharpness, and generally less trueness in the final results.
Create a shapefile delimiting the region of interest for the processing. Ensure that the input imagery fully covers the selected area.
If required, set the desired output GSD (Ground Sampling Distance).
Let SURE select the proper tile size for the processing. Sure chooses a tile size that ensures that the processing stays within the hardware recommendations - typically varying between 2000 and 5000 pixels affecting memory usage and processing speed. For custom tiling exceeding this, we recommend retiling with 3rd party software.
If large water bodies are present in the surveyed area, create a 3D shapefile delimiting all relevant water bodies.
Use distributed processing. Using distributed processing, even when a single machine is in use, gives you the following possibilities:
Add nodes to the processing on demand if necessary.
Keep disk space consumption within the hardware requirements.
Try to have only one flight session per project. For multiple flight sessions, make sure that the projects overlap at least by one strip. After producing the True Orthophoto, blend and adjust the results in a post-processing step, e.g. with OrthoVista. You can also use the region of interest definition if you do not want to produce overlap.
Apply your color balancing and adjustments before processing with SURE. This will ensure optimal texture for True Orthophotos, Point Clouds and Meshes. Global adjustments on the imagery (contrast, brightness, gamma correction) will have no impact on the matching quality. If you use local adjustments (such as local contrast enhancements, compression or conversion to 8 Bit) consider using separate matching and texturing images.
DSM & True Ortho Production
If you have a camera capturing both Nadir and Oblique images, create a project containing only the Nadir images. Restricting the imagery to the Nadir frames speeds up processing and guarantees True Orthos with maximum fidelity (edge & detail sharpness due to optimal observation angle and image orientation), as height information for every raster pixel is extracted. When producing a 3D Mesh additionally, set up an individual project (see remarks below) as this workflow achieves the best possible balance betweeen throughput and quality.
For optimal results, we recommend a forward overlap of 80 percent. Use a sideward overlap of 30 percent for rural areas and a minimum of 60 percent for urban areas. See more details here - also adapting to various cameras.
LiDAR point clouds improve the completeness of the results, in particular for low-overlap flights.
From the DSM, also a DSM Mesh (2.5D) can be generated.
3D Mesh Production
Use all the imagery available. In particular, it is essential to use all available imagery for generating high quality textured facades and vertical structures. The SURE Mesh engine will automatically use the required images and resolutions to produce a maximum quality 3D Mesh while maintaining high throughput. When producing a True Ortho additionally, set up an individual project (see remarks above) as this workflow achieves the best possible balance between throughput and quality.
LiDAR point clouds and LiDAR trajectory information improve the completeness and quality of the resulting 3D meshes.