fix(chatt): register the aerial drape N-S so buildings stop floating over the river#223
Merged
Merged
Conversation
…ArcGIS returns the exact bbox The buildings floated over the river because the NAIP drape was N-S mis- registered against the OSM/terrain layers. Root cause found by TDD, not by eye: ArcGIS exportImage returns the requested bbox EXACTLY only when the requested pixel aspect equals the bbox aspect IN THE REQUEST SR. We requested SR 4326 (degrees) with a metre-proportional (cos-lat corrected) 729x2886 size. Because the degree aspect (0.307) != the metre aspect (0.253 = cos35 * 0.307), ArcGIS expanded the LATITUDE extent to keep pixels square: it returned lat 35.00223..35.06557 instead of the requested 35.0078..35.06 — ~616 m of over-scan on EACH end. The runtime maps row->lat linearly across the assumed box range, so every N-S feature was shifted/scaled and the south-bank (Ross's Landing) buildings landed on aerial pixels that read as open water. Longitude was unaffected (dLon=0), which is why streets registered E-W perfectly and the bug looked purely like a vertical float. Fix: size the drape at the DEGREE aspect (fix E-W resolution at groundWm/mpp, derive height = width * degLat/degLon -> 729x2378). Verified live: ArcGIS now returns ymin=35.0078, ymax=35.06 with 0.0 m over-scan. Re-fetched the drape and re-baked. Building-footprint overlay on the corrected aerial confirms the riverfront buildings now sit on land, not in the river. Registration uses groundWm/groundHm (true metres), not the drape pixel count, so the taller-in-pixels image still maps correctly; manifest.drape now reports the real 729x2378 dims. Tests (fail-first, offline in CI): - fetch-drape: pixel aspect must equal the bbox aspect in the request SR (the no-extent-expansion invariant that was violated). - build-scene: each way-based hero registers on its true OSM footprint centroid (<=5 m) — guards hero placement against the aerial ground truth. Refs #222 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This was referenced Jul 7, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #222 (drape N-S registration). Follow-up to #221.
The real bug (found by TDD, not by eye)
#221 clipped OSM geometry to the box and I mistook "streets contained in the drape extent" for "layers registered." They weren't. Buildings still floated over the river.
Ground-truth investigation (overlaying streets/buildings/heroes on the actual
drape.jpg, then looking) showed:exportImagereturns the requested bbox exactly only when the requested pixel aspect equals the bbox aspect in the request SR. We requested SR 4326 (degrees) with a metre-proportional729×2886size. Degree aspect (0.307) ≠ metre aspect (0.253 = cos35·0.307), so ArcGIS expanded the latitude extent to keep pixels square — it returned lat35.00223..35.06557instead of the requested35.0078..35.06: ~616 m over-scan on each end. The runtime maps row→lat linearly over the assumed box, so every N-S feature shifted and the south-bank (Ross's Landing) buildings landed on pixels that read as open water.Fix
Size the drape at the degree aspect: fix E-W resolution at
groundWm/mpp, deriveheight = width·degLat/degLon→ 729×2378. Verified live: ArcGIS now returnsymin=35.0078, ymax=35.06, 0.0 m over-scan. Re-fetched the drape, re-baked, and confirmed by building-footprint overlay that riverfront buildings now sit on land.Registration uses
groundWm/groundHm(true metres), not pixel count, so the taller-in-pixels image still maps correctly;manifest.drapenow reports the real729×2378.Tests (fail-first, run offline in CI)
Before / after
Building-footprint overlay on the aerial: before — a cluster of footprints floating in the open river; after — footprints stop at the south bank, every one on a real building. (Verified by looking, per the project's screenshots-are-truth discipline.)
All 16 bake/world/stage test files (46 tests) green; type-check clean.
🤖 Generated with Claude Code