You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Keep matched-star photometry available when shape classification is inconclusive. Ignore unresolved single-pixel detections when counting failed shape measurements.
Follow coherent, timestamped stellar-signal trends and recent star/background levels. Normal gradual changes no longer become accumulated losses against the start of the night.
Reject sudden matched-signal loss with fewer stars without requiring a brighter sky. Freeze reference learning during suspicious changes, including rapid multi-frame attenuation.
Add Minimum Session Signal, default 40% of the initial matched reference and configurable from 5–80%. Signal limits include a measured-scatter allowance.
Keep clean references up to six hours through rejected intervals, with bounded trend extrapolation. References older than two hours cannot authorize a star-count recovery.
State why guide recovery was not established, including verified regions and borderline eccentricity.
Add session report locations: default local folder, a chosen folder, or beside the LIGHT folder. Keep the active session in one location; use the new location from the next session. Show the actual folder and any fallback or write failure in Options.
Show photometry separately from unavailable shape evidence in the public and OpenAstro inspectors.
Validate against the September 20 HOO session and the earlier QUAD/shape regressions. See field results.