Files
aimsight-labeling/auto/analyses/SNCF01_2026-10-22T1030_2026-11-01T0000/report.md
T
maximeandClaude Opus 5.5 ddb17c2189 auto: labels requalified under the 5-noise rules (only what stands out from the noise): 32 analyses refined, the 6 paused periods verified, the long labels requalified with the fixes the period analysts found; 1153 -> 961 labels
- 32 analyses: each label kept, rewritten or removed by its analyst (section Requalification of the report; mm magnitudes where the targets have them)
- SNCF01 x3, WILDSTRUBEL 2026-09-01 -> 2026-10-06 x3: verified from their existing labels under the new rules
- long labels (no analysis): auto/analyses/long-labels/<STATION>/; NORDEND-169 and -195 run to 07/11, so NORDEND-283 and -284 (their continuation) are removed
- reading-rules.md, stations.md: the citations of removed or rewritten labels point to their successor, or to [expert]

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 23:58:00 +01:00

8.3 KiB
Raw Blame History

SNCF01: checking the labels of 22–31 October

You asked me to check the 28 existing labels of 22/10 10:30 → 01/11 00:00 against the new rules: keep, correct, remove, add.

In short

13 labels kept, 12 corrected, 3 removed, 1 added (26 labels now). The curves cannot be trusted by day on sunny days (heat shimmer), during four train passages, on the evening of 26/10, or after the camera step of 31/10. They also follow a daily camera cycle that the correction only partly removes. I see no movement of the wall. The camera's step on 31/10 needs watching.

Verification

Kept: 13 · Corrected: 12 · Removed: 3 · Added: 1.

  • SNCF01-71: description. On 22/10 the corrected curves do keep the camera's daily cycle: they stay above the level of the evening before. The old text said the correction absorbed it.
  • SNCF01-72: removed. There is no shimmer on 22/10. The day was overcast, and only one point passes the threshold, at the evening return of the camera.
  • SNCF01-74: end moved 15:00 → 13:15, description. The jumps stop when the sky clouds over. The list of targets was copied from another day; I replaced it with the targets that move that day.
  • SNCF01-76: description. P7 does jump during this passage (about 0.3 mm). The old text said the other targets did not move.
  • SNCF01-77: description. The curves do not come back that evening (the evening episode, then the outage). The old text, copied from another day, said they did.
  • SNCF01-79: start moved 14:30 → 15:15, the first measurement after the common gap, where the drift begins. Description: the targets that drift are named.
  • SNCF01-81: 07:15–16:15 → 08:00–14:45. The jumps on the curves run from 08:15 to 14:30.
  • SNCF01-86: 08:00–16:00 → 10:45–15:30. Nothing passes the threshold before 11:00.
  • SNCF01-87: start moved 08:30 → 07:45, the morning turn of the camera. The text had a duplicated phrase.
  • SNCF01-90: 08:30–19:00 → 09:15–15:30. The shimmer stops in mid-afternoon. The old end took in the evening trains, which have their own labels.
  • SNCF01-93: description. The targets that jump at 21:30 are P3, P1 and BG.3, in x. The old text named P1, P2 and BG.2.
  • SNCF01-94, SNCF01-95: removed. F4.2 is rejected for one measurement and nothing else moves: a missing point is not labelled.
  • SNCF01-96: 07:45–15:45 → 08:15–16:00. Description: P2 and P1 are among the targets that move most.
  • SNCF01-97: description. On 31/10 the curves do not come back at night. Part of this comes from the camera step, which now has its own label.
  • Added: 31/10 12:15 → 12:45, Displacement, the camera step.
  • Labels that cross the range (SNCF01-1 to -7): they hold inside the range. SNCF01-7, the slow drift over the month, does not mention the jump of 31/10 at 12:30. That jump now has its own label.

What happened

The station was often silent. No measurements from 22/10 15:30 to 25/10 08:30, nor from 26/10 22:00 to 28/10 07:15. Its power metrics are empty, so the cause is not known. Short common gaps on 26/10 and 29/10 also hit the other stations.

Daily camera cycle (every day). In the sun the camera turns a little and its image scale grows. The correction removes most of this, but not all. P5, P6 and P7 move by up to about 1 mm during the day and come back in the evening. The cycle is smaller on overcast days (22/10, 26/10). This is likely the camera heating, but there is no temperature on this station to confirm it. Nothing to do about the wall.

Heat shimmer on sunny days (25, 28, 29, 30, 31/10). From mid-morning to mid-afternoon the spots tremble. Single points jump and come straight back, by up to about 2 mm on CV.4 and CV.2 and less on the P targets. At night, and on the overcast days, the curves stay calm. The cause is deduced from the sun on the images; I am fairly sure of it.

Noon images, 22/10 to 31/10 Overcast on 22/10, 25/10 at 12:00 (sunny earlier) and 26/10; hard sun shadows on the wall on the other days.

P5, P6, P7, CV.4 in mm, 25/10 to 31/10 The slow daily waves of P5, P6, P7 (the camera cycle) and the daytime spikes of CV.4 (the shimmer).

Trains and passages. On 28/10 at 21:15 and 22:45, and on 30/10 at 18:00 and 18:30, a passing train hides most targets and two of the three reference targets. For one measurement the remaining points are wrong by up to about 5 mm (BG.3, P1–P4). One train is seen on the 18:00 image. Smaller passages occur on 25/10 at 23:45 and 30/10 at 21:30, and on 28/10 at 10:30 CV.2 locks onto another spot. These are single points to ignore.

Evening of 26/10. From 15:15 to the 22:00 outage, P5 and P6 drift by about 1 mm while others drift the opposite way. At the same time two reference targets (R2, BG.3.1) are seen less well. The night images show nothing clear. The cause is unknown; reduced visibility is likely. By 28/10 the curves are back at their level.

Camera step on 31/10 at 12:30. In full sun, the camera suddenly tilts more than in its usual daily cycle. Every target moves with it, so the wall did not move. The correction leaves a step of about 0.3 mm on P6, P7 and P3. The camera does not return to its earlier night position, and P5 and P6 stay higher until the end of the range. At 17:00 the three reference targets also shift slightly against each other. The cause is unknown; the sun on the camera or its mount is possible.

corr_y in mm, 31/10 The step at 12:30 (grey band), and P5 and P6 not coming back in the evening.

What to do

  • Check the power and transmission of the station: long outages on 22–25/10 and 26–28/10, and no power metrics are recorded.
  • Watch the camera after 31/10. Check in the next days whether its night position and the night level of P5 and P6 come back. If they do not, check the camera mount and the reference targets.
  • The permanent target problems (BG.1, F4.1, CV.7, F4.4, F4.3, CV.6) remain, as labelled before.

Not established

  • The cause of the 31/10 step: sun on the camera, or a mount that slipped. The following nights and a temperature sensor would settle it.
  • The cause of the evening of 26/10: there is no weather data and the night images do not show it.

Labels

Start (UTC) End (UTC) Family Targets In a few words
22/10 10:30 22/10 13:30 No data all intermittent gaps
22/10 10:30 22/10 15:30 Thermal displacement P5, P6, P7 camera cycle kept by the curves (corrected)
22/10 15:00 25/10 08:30 No data all outage of nearly three days
25/10 08:30 25/10 13:15 Visibility CV.4, F4.2, F4.6, CV.9, P1 heat shimmer, sunny morning (corrected)
25/10 08:30 25/10 22:15 Thermal displacement P5, P6, P7 camera cycle
25/10 23:30 26/10 00:00 Obstruction CV, F4.2, P7 passage, P7 point wrong (corrected)
26/10 08:15 26/10 22:00 Thermal displacement P7, P5 camera cycle, overcast (corrected)
26/10 14:30 26/10 15:15 No data all common gap
26/10 15:15 26/10 22:00 Visibility P5, P6, P4, P1, P3 evening drift, references less well seen (corrected)
26/10 22:00 28/10 07:15 No data all outage
28/10 07:30 28/10 22:45 Thermal displacement P5, P6, P7 camera cycle
28/10 08:00 28/10 14:45 Visibility CV.4, CV.2, P2, P3 heat shimmer (corrected)
28/10 10:15 28/10 10:45 Obstruction CV.2 locks onto another spot
28/10 21:00 28/10 21:30 Obstruction most train
28/10 22:30 28/10 23:00 Obstruction most train
29/10 07:45 29/10 23:15 Thermal displacement P5, P6, P7 camera cycle (corrected)
29/10 10:45 29/10 15:30 Visibility CV.4, F4.2, CV.6, BG.3 heat shimmer (corrected)
29/10 14:45 29/10 15:15 No data all common gap
30/10 08:00 31/10 07:45 Thermal displacement P5, P6, P7 camera cycle
30/10 09:15 30/10 15:30 Visibility CV.2, F4.2, CV.6, F4.6 heat shimmer (corrected)
30/10 17:45 30/10 18:15 Obstruction most train, seen at 18:00
30/10 18:15 30/10 18:45 Obstruction most train
30/10 21:15 30/10 22:00 Obstruction P3, P1, BG.3 passage (corrected)
31/10 07:45 01/11 00:00 Thermal displacement P6, P5, P4, P7 camera cycle, no return (corrected)
31/10 08:15 31/10 16:00 Visibility CV.4, P2, CV.2, P1 heat shimmer (corrected)
31/10 12:15 31/10 12:45 Displacement P6, P7, P5, P3 camera step, not erased (added)