Heron Desktop 2.5 Manual

Global Optimization Settings

Also in this case it's possible to use predefined or user defined Presets to run the process.

Preset

Delete currently selected preset from the list

General settings

• Readers quota (maximum maps in RAM): It allows keeping in RAM memory as much local maps as possible. If there is enough memory to store all of them this value should be the number of local maps of the global optimization.
• Default Global Optimization iterations: Minimum number of iterations required within the ICP process to converge to an accepted result. By running the Optimize process the set value will be used instead of the default one.
• Use gravity correction: Check it to take in account also the gravity direction (information stored by the IMU platform) in the registering process. The final optimized 3D model will be oriented according to the vertical direction. 
  It's suggested to run almost the whole Global Optimization process (or at least a first run) without checking it, so as not to insert too many constraints that could limit the global alignment, and then run an Optimize process checking "Use gravity correction". It's very useful during registration between different Heron acquisitions.
  Don't use it if the Global Optimization includes georeferenced cloud constraints (in which the vertical is therefore already known and well defined).
• Gravity confidence (deg): How much do you trust the IMU gravity estimation. In the end this value specifies how much do you want to enforce gravity alignment with respect to the matches that have been selected.
  A value of 0.1 is typically used.
• Constraints confidence
  Low, Medium and HIgh preset are available.

Batch selection settings

• Maximum distance for valid matches (m): Maximum search distance (in meters) to detect a valid match. The shorter the distance, the less matches are searched for.
• Minimum overlap for potential matches (%): Minimum value (in percentage) of the overlap between two local maps to consider a potential match. The bigger the overlap, the faster the process is, because it only looks for matches with a good overlap.
• Minimum distance for global matches (#edges): Minimum distance (in terms of connections -edges- number) to detect a global match. The shorter the distance, the less matches are searched for.

Valid match: a match that has been accepted. It is visualized as an edge connecting two maps (dots) in the 3D view.

Potential match: a match computed between two disconnected local maps that has not been validated yet.

Global match: a match that has been computed without considering the relative poses of the local maps in the current state of the global optimizer.

ICP settings

In this section the parameters can be set for both automatic (Batch) processes and Manual ones separately. We name cloud A and cloud B the two point clouds (local maps or static clouds) included in the ICP process.

• Iterations (#): minimum number of iterations running in the ICP algorithm.
• Correspondences (#): Number of points in cloud A that will be used to retrieve correspondences in cloud B. Only compatible pairs will be used for registration.

      It's  suggested not to significantly increase the number correspondences, because it may slow down too much the selection. Few hundreds (e.g. 400) could be generally be a good upper limit.

• Bidirectional correspondences: If not enabled, N points in cloud A will be randomly selected and their counterpart will be queried to cloud B. If enabled, N/2 points in cloud A will be randomly selected and their counterpart will be queried to cloud B meanwhile N/2 points in cloud B will be randomly selected and their counterpart will be queried to cloud A. This is particularly useful when one local map is very small and the other is very large (only selecting points in the large map would produce a very small overlap, while only selecting points in the small map would produce a 100% overlap).
• Use progressive validation: Check it to enable the progressive refinement; each iterations residuals of correspondences will be sorted and the median value + the final NN max distance/angle will be used as new tolerance. This repeats until, in the last iteration, tolerances are set to the Final NN max dist/angle.

      This is particularly useful when working with very accurate static scans.

• Initial Maximum distance (m): In the first ICP iteration, all correspondences whose points are compatible with the maximum distance will be accepted as valid. 
• Final Maximum distance (m): In the last ICP iteration, tolerances in distances must be lower than this set value.
• Initial Maximum angle (deg): In the first ICP iteration, all correspondences whose points are compatible with the maximum angle will be accepted as valid. 
• Final Maximum angle (deg):  In the last ICP iteration, tolerances in angles must be lower than this set value. 
• Minimum correspondences (#): Minimum number of correspondences  that must be respected to consider the Match as valid.

   Whether the ICP has failed because of few common geometries, it is recommended to manual insert connections between maps by decreasing the number of minimum correspondences below the value of 3.

Useful when merging optimized trajectories with few overlapping areas (e.g. indoor - outdoor).

• Minimum overlap (%): Minimum overlap between local maps that must be respected to consider the Match as valid.

   Whether the ICP has failed because of few common geometries, it is recommended to manual insert connections between maps by decreasing the number of minimum correspondences until the value of 0.

Useful when merging optimized trajectories with few overlapping areas (e.g. indoor - outdoor).

Tips