Class MetricFromUncalibratedPairwiseGraph

java.lang.Object
boofcv.alg.structure.ReconstructionFromPairwiseGraph
boofcv.alg.structure.MetricFromUncalibratedPairwiseGraph
All Implemented Interfaces:
VerbosePrint

public class MetricFromUncalibratedPairwiseGraph extends ReconstructionFromPairwiseGraph
Fully computes views (intrinsics + SE3) for each view and saves which observations were inliers. This should be considered a first pass and all optimization is done at a local level.
  1. Input: PairwiseImageGraph and image information
  2. Selected an initial set of views and perform metric reconstruction.
  3. Grow metric scene one at a time using previously found metric views. MetricExpandByOneView
  4. Stop when all views have been considered
Output is contained in SceneWorkingGraph and accessible from ???. 3D point features are not part of the output directly. Observations used are saved and can be used to triangulate the 3D features. It's advisable to perform bundle adjustment and outlier rejection and the scene as a whole.

Important Note: It's automatically assumed that the image center is the principle point and all pixels are shifted by this amount. This means that the found intrinsic parameters will have (cx,cy) = (0,0).

  • Field Details

    • refineSceneWhileExpandingMaxViews

      public int refineSceneWhileExpandingMaxViews
      When expanding a scene, SBA is applied to the entire scene until it has this many views. This often helps improve the initial metric scene estimate significantly, but can be expensive.
  • Constructor Details

    • MetricFromUncalibratedPairwiseGraph

      public MetricFromUncalibratedPairwiseGraph(PairwiseGraphUtils utils)
    • MetricFromUncalibratedPairwiseGraph

      public MetricFromUncalibratedPairwiseGraph(ConfigProjectiveReconstruction config)
    • MetricFromUncalibratedPairwiseGraph

      public MetricFromUncalibratedPairwiseGraph()
  • Method Details