Prolincur Technologies

DWG / DXFGeoJSON / MVT

DWG and DXF to GeoJSON: putting CAD drawings on the map

How to turn AutoCAD floor and site plans into geo-referenced GeoJSON features and vector tiles, keeping layers, colors and line styles.

Send a sample fileSee case studies

The problem

Why this conversion is hard

  • CAD drawings use local coordinates, usually millimetres or feet from an arbitrary origin, so they do not line up with any map.
  • DWG is a closed binary format. Most open-source GIS tools read DXF, but not every DWG version.
  • Layers, colors, dashed line styles, arcs and text carry meaning that a naive export throws away.

Our approach

How we do it

  1. 01

    Read the drawing

    DXF can be read directly. For DWG we use Autodesk Platform Services, which reads every DWG version reliably.

  2. 02

    Convert entities to features

    Lines, open and closed polylines, arcs and text become GeoJSON geometries. Arcs are tessellated, and layer, color and line type are kept as properties.

  3. 03

    Geo-reference

    A similarity transform (scale, rotation, translation) from two or more control points places the drawing in a projected CRS. Users can fine-tune it by moving and rotating the drawing on the map.

  4. 04

    Serve

    Features are reprojected to WGS 84 and served as Mapbox Vector Tiles, so even large site plans stay fast.

Example

A short working example

A starting point to show the idea. Production pipelines add validation, tiling and error handling around it.

Python: DXF lines and polylines to geo-referenced GeoJSON
import json, math
import ezdxf
from pyproj import Transformer

# Similarity transform from drawing units to a projected CRS (here UTM 43N),
# solved from control points; values below are examples
scale, angle = 0.001, math.radians(12.5)        # mm to m, rotation
tx, ty = 373_500.0, 2_051_200.0                  # drawing origin in UTM
to_wgs84 = Transformer.from_crs("EPSG:32643", "EPSG:4326", always_xy=True)

def to_lonlat(x, y):
    e = tx + scale * (x * math.cos(angle) - y * math.sin(angle))
    n = ty + scale * (x * math.sin(angle) + y * math.cos(angle))
    return to_wgs84.transform(e, n)

doc = ezdxf.readfile("site-plan.dxf")
features = []
for e in doc.modelspace().query("LINE LWPOLYLINE"):
    pts = [e.dxf.start, e.dxf.end] if e.dxftype() == "LINE" else list(e.get_points("xy"))
    closed = e.dxftype() == "LWPOLYLINE" and e.closed
    coords = [to_lonlat(p[0], p[1]) for p in pts]
    geom = {"type": "Polygon", "coordinates": [coords + coords[:1]]} if closed \
        else {"type": "LineString", "coordinates": coords}
    features.append({"type": "Feature", "geometry": geom,
                     "properties": {"layer": e.dxf.layer, "color": e.dxf.color}})

json.dump({"type": "FeatureCollection", "features": features}, open("site-plan.geojson", "w"))

Pitfalls

What usually goes wrong

  • Check $INSUNITS in the DXF header before assuming millimetres.
  • Blocks (INSERT entities) must be exploded with their own scale and rotation.
  • Text height and rotation need to become label properties, not geometry.
  • Keep the transform parameters with the data so the placement can be edited later.

Case studies

Where we have done this

Related

Services and other guides

Want this done on your own DWG / DXF files?

Our CAD to Map accelerator is fixed scope and fixed price. Start with a free sample file, or talk to an engineer.

Talk to an engineerSend a project brief