Conversation
Copy link
Copy Markdown
Contributor
The previous method is sensitive to priors with extreme covariances, producing unstable RANSAC thresholds.
This PR computes the mean RMS stddev from valid covariances and scale it by sqrt(chi-square 95% quantile, dof=3)
to yield an approximate 95% 3D confidence radius.
StonerLing and others added 2 commits
October 21, 2025 13:45
Copy link
Copy Markdown
Contributor
Last request: Can we adjust existing tests or create a new one to cover this new feature? Thanks.