Fundamental Matrix

RANSAC Problem

FundMatProblem is a type alias for MinimalSolveProblem{FundMat{T}, 7, T, ...} — a 2D-2D correspondence problem with the 7-point minimal solver and, for LO refinement, Levenberg–Marquardt on the 7-dof Bartoli–Sturm chart about the current F (signed Sampson residual).

VisualGeometryToolkitEstimation.FundMatProblem — Type
FundMatProblem(pairs; sampler=nothing, backend=NativeSolver())

Construct a FundMatProblem from 2D-2D point correspondences. See HomographyProblem for the input container, its :score prior and sampler.

source
FundMatProblem(pair::PoolPair{<:FundMat}; kw...) -> FundMatProblem

The pool-frame fundamental problem of a pool record: its correspondences (NormalizedImageFrame) verbatim. Its sample gate reads the scale-free tolerance on its normalized view, so it takes no coincident_tol. kw go to FundMatProblem.

source

Fitting Pipeline

VisualGeometryToolkitEstimation.find_fundmat — Function
find_fundmat(correspondences; scoring,
    max_trials, min_trials,
    local_optimization_strategy, config=nothing)

Robust fundamental matrix estimation via 7-point RANSAC with MSAC scoring and least-squares-on-the-inliers local optimization. Every default is configs/default.toml's ([ransac], [pipelines.fundmat]); a keyword overrides.

Returns a RansacEstimate, the :no_model result (iterations 0) for fewer than 7 correspondences. iterations(result) is the number of trials (hypotheses drawn); test found_model(result) before reading the model (see RansacEstimate).

Arguments

  • correspondences: as find_homography; a Correspondences:score column is the per-match prior (PROSAC; uniform otherwise).
  • scoring: the scorer, its width in PIXELS, as find_homography; default [pipelines.fundmat] scoring.
  • workspace: as find_homography, built with RansacWorkspace([T,] FundMatProblem{Float64}, n_data).
  • local_optimization_strategy: default [pipelines.fundmat] local_optimization_strategy, as find_homography.
  • the other keywords as find_homography.
source
VisualGeometryToolkitEstimation.refine_fundmat — Function
refine_fundmat(F, correspondences; optimization_strategy, scale | width, workspace = nothing)
    -> FitResult{M, RefineMetadata}

The fundamental matrix F refined against correspondences from F itself (the Sampson residual, pixels; codimension 1): the keywords, the result and the refusals are those of every refine_*, documented on refine_homography (default strategy: [refine.fundmat] optimization_strategy). Fewer than 7 correspondences: the guess back unrefined, record (ReturnCode.InitialFailure, NoRule) (retcode_symbol:degenerate).

source