Introduction:

Aerotriangulation (AT) is a critical step in photogrammetry that determines the precise orientation and position of images (poses) and calculates the 3D positions of tie points that connect overlapping photos. Selecting the appropriate AT strategy in iTwin Capture Modeler is essential for achieving accurate and reliable results tailored to your dataset and project requirements.

This guide provides practical advice to help you choose the right strategy for pose and tie point settings in iTwin Capture Modeler, based on various use cases and the nature of your data.

 

Understanding the AT Options in iTwin Capture Modeler:

When configuring the AT process in iTwin Capture Modeler, you have four pose policies to choose from:

Compute

Computes the poses (camera positions and orientations) and tie points from scratch, ignoring any existing pose information.

Use When: Pose information is missing, incomplete, or unreliable.

Adjust:

Adjusts the current poses based on the computed tie points. It refines the existing poses rather than recalculating them entirely. This option is only available if all poses are already complete (i.e., have both position and orientation).

Use When: You have initial pose information that is acceptable (not necessarily highly accurate but not wrong) and may need refinement.

Extend:

Computes poses that are not yet part of the main component, computes tie points, and adjusts all poses. It extends an existing AT result by incorporating additional images or merging disconnected components.

Note: Extend is not available for the first AT run.

Use When: You want to add new images to an existing AT result or merge separate AT results.

Lock:

Keeps existing poses and tie points unchanged, without any adjustments or computations.

Use When: You are satisfied with the current results and do not want any modifications.

 

General Guidelines for Aerotriangulation:

Optical Properties:

• Always start with proper input optical properties, either from a complete entry in the database (including distortion parameters) or from a previous result on another dataset or a subset of the current one.
• Activate pre-calibration only if the initial optical properties are unknown or unreliable.

Pose Information:

• If some pose information is incomplete or missing, the only pose policy options are Compute or Extend.
• If initial pose information is acceptable (not wrong), it is worthwhile to try the Adjust pose policy.

Ground Control Points (GCPs):

• Use GCPs whenever possible to improve the accuracy of the AT results.
• When using pre-calibration, having external adjustment constraints like GCPs or GPS data is beneficial.

Improving Results:

• If the accuracy of a first result can be increased, you can run an additional "Adjust poses" AT on top of the initial result.

 

 

Use Case Guidelines in iTwin Capture Modeler:

 

1. Nadir Flight with Complete Pose Information and Proper Initial Optical Properties

• Positioning Strategy: Use position metadata as adjustment constraint.
• Pose Policy: Adjust.
• Optical Settings: Use initial optical properties without pre-calibration and enable adjustment.
• Additional Options:

• • Define the block type as "Vertical views only" and set minimum and maximum viewing distances.
  • Set pairs selection to Exhaustive only if the viewing distances are defined.

Explanation:

With complete pose information (positions and orientations) and accurate optical properties, using Adjust refines your existing poses. Setting position metadata as an adjustment constraint helps prevent large-scale deformations like the "curving" effect in extensive datasets. Defining the block type as "Vertical Views Only" and setting viewing distances optimize tie point matching. The Exhaustive pairs selection improves matching efficiency but should only be used when viewing distances are defined to avoid unnecessary computations.

 

2. Nadir Flight with Position Information and Proper Initial Optical Properties

• Positioning Strategy: Use position metadata as adjustment constraint.
• Pose Policy: Compute.
• Optical Settings: Use initial optical properties without pre-calibration and enable adjustment.
• Additional Options:

• • Define the block type as "Vertical views only" and set minimum and maximum viewing distances
  • Set pairs selection to Exhaustive only if the viewing distances are defined.

Explanation:

Without orientation data, you cannot use Adjust. Compute recalculates poses while the position metadata guides the computation. Accurate initial optical properties enhance the result. Using position metadata as an adjustment constraint reduces the risk of deformation. Setting the block type and viewing distances helps optimize processing. The Exhaustive pairs selection should only be used if both viewing distances are defined and the block type is set to "Vertical Views Only" to ensure efficiency.

 

3. Nadir Flight with Complete Pose Information and Poor Initial Optical Properties

• Positioning Strategy: Use position metadata as adjustment constraint.
• Pose Policy: Adjust.
• Optical Settings: Activate pre-calibration and enable adjustment.
• Additional Options:

• • Define the block type as "Vertical views only" and set minimum and maximum viewing distances.
  • Set pairs selection to "Exhaustive" if viewing distances have been defined.

Explanation:

When optical properties are poor or unknown, activating pre-calibration allows the software to estimate camera parameters during AT. With complete pose information, Adjust refines the poses while pre-calibration improves optical settings. Using position metadata as an adjustment constraint helps prevent curving effects. As in Use Case 1, defining the block type and viewing distances, and cautiously using the Exhaustive pairs selection, optimizes tie point matching.

 

4. Nadir Flight with Position Information and Poor Initial Optical Properties

• Positioning Strategy: Use position metadata as adjustment constraint.
• Pose Policy: Compute.
• Optical Settings: Activate pre-calibration and enable adjustment.
• Additional Options:

• • Define the block type as "Vertical views only" and set minimum and maximum viewing distances.
  • Set pairs selection to "Exhaustive" if viewing distances have been defined.

Explanation:

Lacking orientation data and with poor optical properties, you need to Compute poses and use pre-calibration. Position metadata as an adjustment constraint aids in achieving accurate results and prevents curving effects. Defining the block type and viewing distances, and using the Exhaustive pairs selection cautiously, enhances processing efficiency and accuracy.

 

5. Nadir Flight with Several Inconsistent Groups with Complete Poses and Accurate Optical Properties

• Positioning Strategy: Use rigid registration on position metadata.
• Pose Policy: Adjust.
• Optical Settings: Use initial optical properties without pre-calibration and enable adjustment or keep unchanged.

Explanation:

When multiple groups (e.g., different flights) are inconsistent, using rigid registration on position metadata aligns them without constraining the results to several inconsistent positions. We do not recommend using position metadata as adjustment constraints in this case to avoid enforcing the inconsistencies. Adjust refines poses within each group, ensuring consistency across the dataset.

 

6. Previous AT is Curved or Incomplete

• Positioning Strategy: Use rigid registration on position metadata.
• Pose Policy: Extend.
• Optical Settings: Use initial optical properties without pre-calibration and enable adjustment.
• Process:

• • First, process individual flights separately using settings from Case 1.
  • Merge the individual AT results.
  • Run an additional AT with the Extend pose policy after merging.

Explanation:

Processing individual flights separately can resolve issues like curving or incomplete AT results. After merging, Extend incorporates all data into a refined AT result.

 

7. Nadir Flight with Several Inconsistent Groups and Future Global Shift Support

Note: The "global shift" feature is intended for future implementation.

• Positioning Strategy: Use position metadata as adjustment constraint.
• Pose Policy: Adjust.
• Optical Settings: Use initial optical properties without pre-calibration and enable adjustment.
• Additional Options:

• • Apply global shift per photo group (when available).

Explanation:

With inconsistent groups and when global shift support becomes available, you can use position metadata as adjustment constraints to correct for large-scale discrepancies. The global shift feature estimates biases between flights, helping to align them accurately. Using this strategy prevents curving effects and ensures consistent alignment across the dataset. Adjust refines poses within each group.

 

8. Any Dataset with Bad Pose Information

• Positioning Strategy: None.
• Pose Policy: Compute.
• Optical Settings: Activate pre-calibration if optical properties are unknown.

Explanation:

When pose information is unreliable or missing, it's best to compute poses from scratch without using faulty position metadata that could introduce errors.

 

9. Nadir Dataset with Control Points and Position Information Without Bias

• Positioning Strategy: Use control points and position metadata as adjustment constraints.
• Pose Policy: Depends on availability of orientation data:

• • If orientations are available, use Adjust.
  • If not, use Compute.
• Optical Settings: Use initial optical properties; activate pre-calibration if necessary.

Explanation:

When control points and position metadata are consistent (no translation bias), using both as adjustment constraints enhances accuracy. The pose policy depends on whether you have complete pose information.

 

10. Nadir Dataset with Control Points and Discrepancies Between Control Points and Position Information

• Positioning Strategy: Use position metadata as adjustment constraint and rigid registration on control points.
• Pose Policy: Depends on availability of orientation data.
• Optical Settings: Use initial optical properties; activate pre-calibration if necessary.

Explanation:

When there's a discrepancy (e.g., translation bias) between control points and position data, using rigid registration on control points helps reconcile differences. This strategy aligns the dataset based on reliable control points while still using position metadata to guide the adjustment. Adjusting the positioning strategy accordingly improves overall accuracy.

 

11. Nadir Dataset with Control Points and Discrepancies (Future Global Shift Support)

Note: The "global shift" feature is intended for future implementation.

• Positioning Strategy: Use control points and position metadata as adjustment constraints.
• Pose Policy: Depends on availability of orientation data.
• Optical Settings: Use initial optical properties; activate pre-calibration if necessary.

Explanation:

With inconsistencies and future global shift support, you can adjust for discrepancies between control points and position data, improving overall accuracy.

 

12. Complex Structures (e.g., Towers) and Unsatisfactory Previous AT Results

• Positioning Strategy: Use rigid registration on position metadata.
• Pose Policy:

• • Start with Adjust if pose information is acceptable.
  • Switch to Compute if results show issues like duplication or artifacts.
• Optical Settings: Use initial optical properties; activate pre-calibration if necessary.
• Additional Options:

• • Define the block type as "Orbit Around Thin Vertical Object" for towers to improve tie point accuracy in the foreground.

Explanation:

For complex structures like towers, starting with Adjust refines existing poses. If the initial AT results are not satisfactory (e.g., due to GPS challenges or complex geometry), recomputing poses with Compute may resolve issues. Using rigid registration on position metadata aligns the dataset based on position data. Defining the block type as "Orbit Around Thin Vertical Object" enhances the accuracy of tie points, particularly in the foreground.

 

13. Any Dataset with Accurate Pose Information

• Positioning Strategy: As appropriate.
• Pose Policy: Adjust.
• Optical Settings: Use initial optical properties; activate pre-calibration if necessary.

Explanation:

With accurate pose information, adjusting poses refines results without discarding existing data.

 

 

Additional Notes on Specific Scenarios:

 

3D Objects with Medium to Small Extent

• Positioning Strategy:

• • Use position metadata as adjustment constraint only if using high-accuracy positioning (e.g., RTK GPS).
• Pose Policy:

• • Start with Adjust; switch to Compute if results show duplication or artifacts.
• Optical Settings: Use initial optical properties; activate pre-calibration if necessary.

Explanation:

High-accuracy positioning data benefits small-scale 3D objects. Adjusting poses refines the model, but computing may be necessary if issues arise.

 

Two Facing Oblique Flights with Ambiguous Patterns

• Challenges:

• • Ambiguous patterns can cause misalignment (e.g., matching one side of a bridge to the other incorrectly).
• Strategy:

• • Consider splitting the dataset into blocks and processing separately.
  • Perform clean-up iterations (reset bad poses) and use the Extend pose policy after merging.
• Optical Settings: Use initial optical properties; activate pre-calibration if necessary.

Explanation:

Separating ambiguous data helps prevent misalignments. Merging and extending allows for a comprehensive model without confusion between similar patterns.

 

Bridges and Structures with GPS Challenges

• Challenges:

• • GPS accuracy may degrade near bridges due to signal interference.
  • Low overlap between nadir, oblique, and ground photos.
  • Large viewpoint changes and ambiguous patterns.
• Strategy:

• • Start with Adjust; if results are poor, switch to Compute.
  • May need to process nadir and oblique flights separately.
• Positioning Strategy:
• • Use rigid registration where appropriate.
• Optical Settings: Use initial optical properties; activate pre-calibration if necessary.

Explanation:

Structures like bridges present unique challenges due to GPS signal interference and complex geometries. Begin by adjusting poses using Adjust. If issues like duplication or misalignment occur, switch to Compute. Processing different flight types separately (nadir, oblique) and then merging can improve results. Adjusting strategies based on results ensures better accuracy.

 

Best Practices:

• Define Viewing Distances:

• • When initial positions are available, define minimum and maximum viewing distances to improve tie point matching.
  • Avoid setting pairs selection to Exhaustive if initial poses are unknown or viewing distances are undefined, as it may lead to excessive computations and potential mismatches.
  • Only one setup for the whole block so more suitable for nadir flights with close to constant height
• Use Ground Control Points (GCPs):

• • Incorporate GCPs to enhance the georeferencing accuracy of your model.
  • Ensure that GCPs and position metadata are consistent to avoid introducing biases.
• Iterative Refinement:

• • It's often beneficial to run multiple AT processes, adjusting settings based on previous results.
  • If initial results are unsatisfactory, consider adjusting pose policies or optical settings before reprocessing.

 

Conclusion:

Selecting the appropriate aerotriangulation strategy in iTwin Capture Modeler is essential for achieving precise and accurate 3D models. By understanding the nature of your dataset and following these guidelines, you can choose the right pose policy and settings to optimize your results.

Key Takeaways:

• Start with Good Data: Ensure optical properties and pose information are as accurate as possible.
• Use Constraints Wisely: Apply position metadata and control points as adjustment constraints when they are reliable.
• Be Flexible: Adjust your strategy based on the completeness and quality of your data.
• Iterate and Refine: Don't hesitate to run multiple AT processes to improve accuracy.
• Seek Assistance: For complex cases or uncertainties, reach out to our support team or consult with experts.

By carefully selecting your AT strategy, you can maximize the quality of your photogrammetric projects using iTwin Capture Modeler.