R&D

From Report to Map — Automated Drill Hole Extraction

A 554-page NI 43-101 technical report (Lac des Iles Mine, Ontario) parsed automatically: drill hole tables extracted from the PDF appendices, local mine-grid coordinates georeferenced to UTM / latitude-longitude, and rendered below as interactive 2D and 3D views. No manual data entry.

950Drill holes
554PDF pages parsed
8Resource zones
178With azimuth / dip
2D map — collars on satellite imagery

Drill hole collars — WGS84

Coordinates converted from the local mine grid (rotated 359.725° true) to UTM 16N NAD83, calibrated against 157 published collars from the Ontario Drill Hole Database. Positional accuracy approximately ±50 m — adequate for mapping, not for survey work. Click a collar for details; use the layer control to filter zones.

3D view — hole traces

Collars & straight-line traces — mine grid

drag: rotate  |  shift+drag / right button: pan  |  wheel: zoom

All collars plotted at a single datum (no topography yet). The 178 holes with a published azimuth / dip are drawn as straight-line traces to their reported length; the rest appear as collar points only. Custom canvas renderer — no external libraries.

About this experiment

Technical reports lock structured data inside PDF tables. This page is a test of a fully automated pipeline: parse → georeference → visualize, from a public NI 43-101 filing to an interactive map, with every step scripted.

Pipeline

  • Extract: drill hole listings (ID, XYZ, length) parsed from the Section 14 appendices of the 2017 Lac des Iles feasibility study.
  • Georeference: the mine grid is local. Rotation from the documented grid bearing + UTM convergence; translation fitted by robust least squares against 157 matching collars in the Ontario Drill Hole Database.
  • Enrich: azimuth and dip recovered for 178 holes by cross-matching the public database (hole ID + length).

Caveats

  • The report publishes no per-hole azimuth/dip — orientation exists only where the public database overlaps (mostly 2000–2001 campaigns).
  • Reference collars are digitized and noisy (±100 m scatter); the fitted transform is good to roughly ±50 m systematically.
  • 3D view uses a single collar elevation — topography not yet applied.

Sources