Four positive-depth counts; a tied maximum cannot qualify a unique camera.
Operational trial target only; deterministic finite-generator proposals do not certify physical correctness.
Eighth/largest original-design singular ratio on the entire final inlier population, null without a resolved fit.
Final re-estimations, including repeats after a changed sampling target.
True only when every final fixed-subset optimizer terminates and its final inlier set stabilizes.
Null without four final epipolar inliers to test the planar competitor.
Inlier count divided by all tentative matches.
Fractions of original pixel-chart area enclosed by observed inlier convex hulls, first/second; no surface coverage claim.
Final calibrated epipolar inlier count.
Nonconverged local proposals or final re-estimation failures, including provisional changing subsets. Unexpected runtime errors fail the operation instead.
Eligible proposals refined in draw order before winner comparison, bounded by the declared local budget.
Median ray angle for finite positive-depth points in radians, or null when absent.
Final inlier median mean-two-sided point-to-line distance in pixels, or null without finite support.
Proposal models refused by their declared numerical owners.
Final fixed-subset optimizer diagnostics, empty when no final optimizer completed.
Final inlier linear-interpolated 90th percentile point-to-line distance in pixels.
Positive-in-both-cameras count divided by all final inliers, or null without a resolved candidate.
Required ideal all-inlier trials at the final support; null denotes an unattainable finite target.
Actual eight-point proposal count, including numerical refusals.
Final calibrated support and independent camera-qualification diagnostics.