v0.5.9 begins Contour Studio: a post-analysis workflow that turns a correctly analyzed LiDAR scan into an evidence-aware contour deliverable, kept out of the crowded analysis panel. This build lands the first foundation for that workflow together with a batch of scientific-correctness and honesty fixes that stand on their own.

v0.5.8 tied every output to the build, method, and unit assumptions that produced it. v0.5.9 carries that discipline into terrain deliverables and into the validation path behind them.

OpenLiDARViewer remains browser-native and local-first: local files stay on the user's device, and no account is required.

Contour Studio foundation

A pure, tested launch-state core now decides whether the deliverable launcher should be hidden, disabled, offered as exploratory, or offered as a full deliverable, using only facts the analysis already produced. Hard blockers (no terrain surface, no ground source, an area that is mostly unsupported) outrank soft caps (unknown vertical units, a geographic CRS, an incomplete stream, no recommended interval, sparse support). The core carries no UI or rendering dependencies, so it is testable without a browser and keeps the science layer clean.

This build ships the launcher and workspace, not just the core. After analysis, a Terrain Products surface leads the panel with the Contour Studio launcher; opening it reveals purpose presets, a unit-aware review, an evidence ladder, and a single export section that routes to the viewer's existing contour, DEM, and terrain-report exporters. Selecting a purpose regenerates the exported contour geometry at a bounded per-purpose generalization tolerance (Survey Review = exact analytical isolines at tolerance 0; Terrain Research, Engineering Plan, Presentation Map and Custom generalize at increasing tolerance), so the cartographic purposes serialize pairwise-distinct vertices, and it stamps the geometry method, purpose, and exact tolerance into the file's provenance. The evidence resolver is the enforced production caller for these exports: the Contour Studio vector exports (GeoJSON, DXF, SVG), the contour report PDF, the DEM raster package (DTM claim), the complete deliverable, and the terrain intelligence report are all minted through the central contour export permit and evidence resolver — a blocked permit refuses the export and the decision is stamped into provenance. A "Complete (ZIP)" deliverable bundles the contours (GeoJSON), the DTM raster, a multipage technical report PDF, provenance, a README, and a verifying SHA256SUMS. What remains for a complete Contour Studio: the launcher still lives inside the analysis panel rather than a separate app-shell surface.

Validation correctness

Hold-out RMSE re-runs ground classification per fold with the held-out points excluded, removing the classify-before-split optimism. The surface fit was already train-only; the remaining leak was the classification step, and the analyser now passes a train-only reclassifier on the shipped terrain path — the SAME SMRF classifier with the SAME resolved parameters, run on the training points only — so a held-out point never helps decide its own ground membership. The report's full-cloud disclosure is off because the leak is removed, not restated.

Confidence calibration no longer reports its quality on the same samples it was fit on. Reported reliability and Brier score are computed by deterministic K-fold cross-fitting, so no sample is scored by a calibrator trained on it.

Unit honesty

Picked-point elevation stops asserting metres it does not know. A foot vertical datum prints feet, a metre datum prints metres, and an unknown or local vertical scale prints no suffix rather than a fabricated metre value. Space reports no longer label unknown-scale data as "metres (assumed)".

Evidence gate coverage

Measurements CSV, the integrity report, and the map-sheet PDF now route their claim status through the one central evidence gate instead of exporting ungated. Unvalidated products carry the exploratory status honestly; nothing is promoted to validated that the registry does not support.

Registration

Change-detection alignment gains trimmed ICP with a median-based warm-start, so a gross blunder or scattered outliers no longer collapse the fit. Diagnostics report the inlier fraction over the whole cloud and the kept-set residual, without hiding rejected outliers.

What this build does not claim

v0.5.9 is not a survey-certification release, and Contour Studio is not yet a complete, shipped feature.

The Contour Studio launcher, workspace, and purpose-driven exports ship and work, and the central export permit / evidence resolver is the enforced production caller for the vector exports (GeoJSON, DXF, SVG), the contour report PDF, the DEM raster package (DTM claim), the complete deliverable, and the terrain intelligence report. The multipage technical-report PDF and the complete ZIP are wired to the Studio's own emitters, and the hold-out classify-in-fold fix runs on the shipped analyser. What still does not ship: moving the launcher fully outside the analysis panel, and independent cross-implementation / field accuracy (E4+) validation. These are disclosed rather than presented as finished.

Deploy

Static files. Host on GitHub Pages, Netlify, a static CDN, or any conventional web host.

Citing this release

Cite OpenLiDARViewer with the metadata in CITATION.cff:

• Version: 0.5.9
• Release date: 2026-07-15
• License: MIT

When the tagged release is archived on Zenodo, cite the version DOI assigned to that snapshot.

Live demo: https://lidar.aurtech.mx/
GitHub: https://github.com/Aurtechmx/openlidarviewer

Open Source • Open Data • Open Exploration