Stitching Drone or Scanned Map Sheets into One Map: A Practical Guide (Malaysia)

Sooner or later every estate, forestry or survey office ends up with the same problem: the area of interest is bigger than one picture. A drone flight produces hundreds of photos. An older survey comes as a pile of scanned sheets. Either way you need one continuous, georeferenced map you can measure on, print and carry to the field.

The two jobs look alike but work differently. This guide walks through both, with the checks that save you from a map that looks fine and is wrong.

Two different problems

Drone photos are overlapping images, each tagged with a GPS position by the aircraft. Software matches features between photos, works out where each was taken, and produces an orthophoto: a top-down image corrected for camera tilt and terrain. This is called photogrammetry, or structure from motion.

Scanned map sheets are already flat images. There is nothing to reconstruct. The work is georeferencing: telling the GIS where each sheet sits by matching known points on the sheet to real coordinates, then joining neighbouring sheets so the edges line up.

Knowing which one you have decides your tools and your error sources.

Stitching drone photos: what decides the result

Overlap. Plan roughly 75 to 80 percent forward overlap and 60 to 70 percent side overlap for mapping flights. Too little and the software cannot find enough matches; the result has holes or warped patches. Flat, repetitive ground such as oil palm canopy or open water is the hardest case, so lean to the high end there.

Consistent conditions. Fly at one height, in steady light, in one session where possible. Mixed heights or a cloud shadow moving across the block create visible tone changes and can confuse the matching.

Sharp images. Motion blur from flying too fast kills matches. Check the first few photos before you fly the rest.

Positions. Ordinary drone GPS is good enough for a map you look at, not one you survey from. Absolute accuracy is typically in metres. Reaching centimetres needs RTK/PPK positioning or ground control points (GCPs), which are surveyed markers visible in the photos. If you used neither, say so on the finished map.

Source: Pix4D: designing the image acquisition plan.

Stitching scanned map sheets: what decides the result

Scan quality. Scan at 300 dpi or higher on a flatbed or a proper large-format scanner. Phone photos of sheets bend the paper and distort scale.

Control points. Georeference each sheet with at least four control points spread across it, ideally at printed grid intersections rather than at features that may have moved. QGIS's Georeferencer covers this workflow.

Know the sheet's datum. Older Malaysian sheets may be in Kertau 1948 or Timbalai 1948 with an RSO projection, while your GIS project and your phone use WGS84 or GDM2000. Choose the correct source system when georeferencing. Getting it wrong shifts the whole map by tens to hundreds of metres.

Paper moves. Paper shrinks, stretches and folds. A polynomial transform can bend a sheet to fit points, but a heavily bent sheet is a sign of a poor scan or a poor control point. Check the residuals your software reports.

Edges. Neighbouring sheets often overlap by a margin or a collar. Crop to the neatline, then check that a road or boundary continues across the join.

Sources: QGIS: Georeferencer, Timbalai 1948 (EPSG:4298), GDM2000 (EPSG:4920), JUPEM.

Checks that catch bad maps

1. Independent check points. Keep two or three known points out of the process, then measure their error on the result. Checking against the points you used proves nothing.
2. Seam walk. Zoom into every join and follow a linear feature across it.
3. Overlay a trusted layer. Put a known boundary or road over the result and look for a constant offset (a datum problem) versus a growing one (a scale or warping problem).
4. Write the accuracy on the map. A date, the method and an honest accuracy statement stop a reference map from being used as a legal one.

Where Vic Tech's tools fit

Vic Tech builds VicTerra for the drone case. It is a Windows desktop tool that stitches a flight into an orthophoto and prints it under the linework of the GeoPDF task sheet the office already issues, so the boundaries, grid and scale bar stay and this month's ground appears beneath them. The whole job runs on the estate's own computer; photos are not uploaded. The output is a PDF sheet, not a data file, and you can open it, print it or load it into the VicMap phone app. It is reference imagery accurate to roughly ±3 metres, and is not for survey, boundary, area or contractual use.

VicMap is the field side. It opens the finished GeoPDF on Android or iOS, shows your live GPS position on it, and works offline. Start with the VicMap 14-day free trial, or see VicTerra. For VicTerra pricing, ask us directly on Telegram.

VicTerra handles drone photos. For scanned paper sheets, use a GIS georeferencer as described above, then export the result as a GeoPDF and carry it in VicMap.

Frequently asked questions

How much overlap do drone photos need for stitching?
A common guideline for mapping flights is about 75 to 80 percent forward and 60 to 70 percent side overlap. Uniform canopy such as oil palm benefits from the higher end, because the software has fewer distinctive features to match.

Can I stitch scanned map sheets without ground survey?
Yes, if the sheets carry printed grid coordinates. Use grid intersections as control points, and select the sheet's original datum and projection. Ground survey is only needed when the sheet has no reliable coordinates.

Is a stitched drone map accurate enough for boundary or area claims?
Not by itself. Without RTK/PPK or ground control points, absolute accuracy is usually in metres. Use it as reference imagery; formal boundary and area work belongs to a licensed land surveyor.