How to measure stockpile volume with a drone

A stockpile volume from a drone is not a guess from a photograph. It is a measurement of a surface, and once you know what the surface is and what it is being measured against, the number stops being a mystery and starts being something you can check. This guide walks through the four parts: the flight, the surface the flight produces, the base under the pile, and the arithmetic between them.

1. The flight gives you a surface, not a volume

A mapping drone flies a grid over the yard taking overlapping photos, usually with about 75% overlap between neighbouring frames. Photogrammetry software finds the same points in many photos and works out where each one sits in three dimensions. The result is a dense set of points on the ground and on the piles, turned into two files: an orthomosaic, which is the map image, and a digital surface model (DSM), which is a grid of heights, one per pixel. (What the three outputs are, and which to measure on.)

A standard mapping flight at around 80 to 120 m gives a ground resolution of 2 to 3 cm per pixel. A pile 40 m across is then described by one to three million heights. That is the surface. The volume still has to be measured from it.

2. A volume is the surface minus a base

The maths is simple. Draw a boundary around the toe of the pile. For every pixel inside it, take the height of the surface, subtract the height of a base, and multiply by the area of the pixel. Add them up. Pixels where the surface stands above the base are material (reported as cut, because that is what you would have to remove to reach the base); pixels below the base are a hollow (reported as fill). For a pile, nearly everything is cut and the net volume is the pile.

Everything that goes wrong with stockpile volumes goes wrong in the base. The surface is measured; the base is decided, and it has to describe the ground that was under the pile before the pile was there.

3. Choosing the base

The bases survey software offers correspond to the kinds of ground a pile sits on.

Why a flat base reads high on a slope

Suppose a conical pile stands on ground that falls at 1 in 20, a slope you would barely notice walking across it. Measure it from the lowest point inside the boundary and the base is a flat plane through the downhill toe. Uphill of that, the wedge of hillside between the plane and the real ground is counted as pile. In our own test of exactly this case, the cone came out 68% high. Measured from a plane fitted to the ground along the boundary, the same cone came out within about one percent of its formula volume.

The lesson is not that the flat base is wrong; it is right on a flat yard. The lesson is that the base should be printed with the result, so whoever reads the number can see what it assumed. Clarity shows the base, the area measured and the height range beside every volume for that reason.

Where to draw the boundary

Draw on the ground just outside the toe, not on the pile. A best-fit plane is fitted to the heights along the line you draw, so a boundary that climbs part-way up the heap tilts the base into the material and the pile reads low. Keep the boundary off anything that is not ground: a loader, a wall, the next pile. If the base cannot be found everywhere inside the boundary, the software should say how much of the area it measured; a volume that quietly covers 80% of the pile is a wrong number with a confident face.

4. A worked example

A pile of crushed stone, 38 m by 31 m at the toe and 6.2 m high at its peak, flown at 100 m with a 20 MP camera (about 2.7 cm per pixel). The boundary encloses 925 m², or roughly 1.27 million pixels.

• Mean height of the surface above a best-fit plane across the boundary: 2.84 m.
• Volume = 2.84 m × 925 m² = 2,627 m³ (3,436 yd³).
• At a loose density of 1.6 t/m³: 4,200 t.
• From the lowest point instead, on ground falling 1 in 25 across the pile: 3,310 m³, 26% more.

The second line is the whole measurement; the software does it for a million pixels instead of one mean. The fourth line is why the base matters more than the pixels. For a rough figure from a tape measure before anyone flies, there is a stockpile volume calculator for cone and windrow piles.

5. What limits the accuracy

In rough order of how much they move the number:

1. The base. Tens of percent, as above. Chosen, not measured; print it.
2. Density, if you want tonnes. Five to fifteen percent between loose, compacted and wet material. A number from the supplier or a measured sample beats a table.
3. The flight's scale. Without ground control, RTK or PPK, the model's scale comes from the drone's own GPS and can be off by up to a few percent, which carries straight into the volume. With any of the three it is usually well under one percent. (What each gives you, in centimetres.)
4. Resolution and overlap. At 2 cm per pixel the surface itself is good to a few centimetres; on a 6 m pile that is under one percent of the volume. Thin overlap and a windy day cost more than resolution does.
5. Vegetation and loaders. A surface model includes whatever stood there when the photo was taken. Fly when the yard is still, and keep machines out of the boundary.

6. Checking a number before you send it

A volume you can defend has three things beside it: the base it was measured from, the fraction of the boundary that was actually measured, and the height range inside, which catches a boundary that strayed onto the next pile. If you have an earlier flight, measure the same pile against it: the difference is what was added or removed, and it does not depend on the base at all. For an inventory that an auditor will read, see stockpile inventory for quarries and aggregates.