GeoRefMaps

Semi-automatic georeferencing for geological maps and scanned figures

Extract coordinate labels from PDF or raster borders, review four corner controls and export a documented GeoTIFF without hiding CRS assumptions.

Warning

Automatic text detection proposes control points; it does not prove that the map datum, UTM zone or corner assignment is correct. Review every corner and compare the result with a trusted reference layer before using it analytically.

Workflow at a glance

PDF page / scanned map / raster image
                  │
                  ▼
 Vector text extraction ──or── border-focused OCR
                  │
                  ▼
 Coordinate candidates + grid-line matching
                  │
                  ▼
 Four editable corner controls + explicit source CRS
                  │
                  ▼
 JSON GCP record ──► georeferenced/reprojected GeoTIFF

Capabilities

Area	Current support
PDF	First-page rendering and vector-text extraction with PyMuPDF
Raster	PNG, JPEG and TIFF image input
OCR	Tesseract fallback focused on map-border regions
Coordinates	Decimal/DMS geographic labels and projected easting/northing candidates
Matching	Corner templates and raster grid-line assistance
Review	Manual correction of proposed corner values and modes
Export	JSON control-point record and RGB GeoTIFF
Reprojection	Optional output EPSG through Rasterio

Supported inputs

• PDF (.pdf), currently processing the selected/first page;
• PNG (.png);
• JPEG (.jpg, .jpeg); and
• TIFF (.tif, .tiff).

Text-based PDFs are preferred because their labels can be extracted without OCR. Scanned figures require a working Tesseract installation and sufficiently clear border labels.

Spatial-reference rules

The current source-CRS presets are deliberately limited:

Coordinate mode	Source CRS
Geographic longitude/latitude	WGS 84 — EPSG:4326
UTM 17 South	WGS 84 / UTM zone 17S — EPSG:32717
UTM 18 South	WGS 84 / UTM zone 18S — EPSG:32718
UTM 19 South	WGS 84 / UTM zone 19S — EPSG:32719

Maps in PSAD56, SAD69, local mine grids or another datum must not be forced into these presets. Their controls require an explicit transformation workflow that is not yet generalized in this MVP.

Installation

git clone https://github.com/jordan-zav/GeoRefMaps.git
cd GeoRefMaps
python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
python -m pip install -r requirements.txt

Install Tesseract OCR separately when scanned maps need OCR. Vector-text PDF processing does not require it.

Run

python -m app.main

On Windows, run_georefmaps.bat provides the same local launcher.

Operator checklist

1. open the source PDF or raster and inspect its stated datum/CRS;
2. review detected coordinate candidates and the inferred mode;
3. correct the four corner assignments manually where necessary;
4. choose the proper UTM zone or geographic mode;
5. export the JSON record before generating the GeoTIFF;
6. reopen the GeoTIFF in QGIS and verify CRS, extent, rotation and residual fit;
7. overlay roads, grids or other trusted features before interpretation.

Outputs

Product	Purpose
GCP JSON	Preserve coordinate mode, source CRS and reviewed corner controls
GeoTIFF	RGB georeferenced map using the accepted corner frame
Reprojected GeoTIFF	Optional resampling into a user-selected destination EPSG

The current four-corner transform is appropriate for rectangular map frames. It is not a replacement for a higher-order GCP adjustment when the source is warped, rotated irregularly or distorted by scanning.

Repository map

Path	Contents
app/main.py	Tkinter interface and operator workflow
app/pdf_processing.py	PDF/raster loading, rendering and vector-text extraction
app/ocr.py	Tesseract discovery and border-focused OCR
app/detector.py	Coordinate parsing, candidate scoring and corner matching
app/export.py	JSON, GeoTIFF and reprojection output
app/models.py	Shared control-point and detection models
assets/branding	Project visual identity

Project status

GeoRefMaps v1.0 is an MVP optimized for rectangular geological map frames from Peru with WGS84 geographic or UTM 17S–19S labels. It has no automated regression suite yet. Broader datum handling, multipage selection, rotated/nonlinear GCP adjustment and quantitative residual reporting remain future work.

License and contact

GeoRefMaps is distributed under the GNU General Public License v3.0.

Jordan Zavaleta — GisGeo Dev
jordanzav@gisgeo.dev · gisgeo.dev