How drone photogrammetry measures the ground
Photogrammetry reconstructs geometry from overlapping photographs. The drone flies a grid and takes a photo every few meters, with each point on the ground appearing in many images. Processing software finds the same features in several photos, solves for the exact position and orientation of every camera (a step called bundle adjustment), and then triangulates millions of matched pixels into a dense point cloud.
From that point cloud come the standard deliverables: a digital surface model (DSM), a digital terrain model (DTM) after ground classification, an orthomosaic in which every pixel is true to scale, a textured 3D mesh and contours. Because the geometry and the imagery come from the same photos, the orthomosaic and the elevation model line up perfectly.
The resolution is set by ground sampling distance (GSD), the size of one pixel on the ground. GSD depends on flight altitude, focal length and sensor pixel size. A common rule of thumb is that horizontal accuracy with good ground control is often quoted at 1 to 2 times the GSD, and vertical accuracy at 1.5 to 3 times the GSD. These are general guidelines, not guarantees, and they assume good overlap, texture and control.
How drone lidar measures the ground
Lidar (light detection and ranging) measures distance directly. A laser scanner fires hundreds of thousands of pulses per second and times each return. Combined with a high-grade GNSS receiver and an inertial measurement unit (IMU) that track the sensor position and attitude at every instant, each return becomes a point with known coordinates.
The key property is that a single laser pulse can produce several returns. Part of the beam hits a leaf, part hits a branch, and part reaches the ground. The last return from each pulse often lands on the bare earth under vegetation, and that is the reason lidar exists in the surveying toolkit. Photogrammetry can only see what the camera sees; lidar can see between leaves.
Lidar does not need texture or light. It works on uniform surfaces, in shadow and at dusk. It does need careful calibration: boresight alignment between the scanner and the IMU, strip adjustment between flight lines, and good GNSS geometry during the flight.
Photogrammetry vs lidar under vegetation
This is the clearest difference. Under a closed tree canopy, photogrammetry produces a surface of the treetops. No amount of processing recovers the ground below, because no camera ever saw it. Ground classification can remove scattered bushes and low crops where the ground is visible between them, but dense cover defeats it.
Lidar with multiple returns penetrates gaps in the canopy and measures the ground below. For topographic surveys of wooded corridors, forest roads, drainage under brush or archaeological features under trees, lidar is usually the only drone method that delivers a usable terrain model.
- Open ground, bare earth, cleared construction sites: photogrammetry works well.
- Low crops, grass and sparse shrubs: photogrammetry can work with careful classification and field checks.
- Dense forest, heavy brush, riparian corridors: lidar is the right sensor.
Accuracy: what each method really delivers
Both methods can reach survey-grade results on the right job, and both can produce misleading results on the wrong one. Accuracy is a property of the whole workflow, not the sensor alone.
For photogrammetry, the main drivers are GSD, image overlap, surface texture, camera calibration and ground control. On open, well-textured ground with ground control points and a low flight, vertical accuracy is usually excellent. On uniform surfaces such as fresh snow, water, clean sand or glossy roofs, matching fails and the surface gets noisy.
For lidar, the main drivers are the GNSS and IMU quality, the boresight calibration, the point density and the range to target. Lidar vertical accuracy is consistent across textures, but its horizontal detail is often softer than a low photogrammetry flight, because the laser footprint and point spacing are coarser than a centimeter-scale pixel.
In both cases, the only honest statement of accuracy comes from independent checkpoints: surveyed targets that were not used in processing. Our article on drone mapping accuracy explains how checkpoints and RMSE work for both methods.
Deliverables compared
Photogrammetry produces a photographic record by default. The orthomosaic is often the most used deliverable on a construction or mining site, because people read a photo faster than a point cloud. The textured 3D model is also a natural output.
Lidar produces a point cloud first. Many lidar payloads carry a camera to colorize points, but the imagery is usually a secondary product with lower resolution than a dedicated mapping camera. If your client needs a sharp orthomosaic, a lidar-only flight may need a second photogrammetry pass.
- Orthomosaic: photogrammetry is stronger.
- Textured 3D mesh: photogrammetry is stronger.
- Bare-earth DTM under vegetation: lidar is stronger.
- Power lines, thin wires and fences: lidar captures thin structures that image matching tends to miss.
- Volumes on open stockpiles: both work; photogrammetry is the common choice because of cost and the photo record.
Cost and operational differences
Cost is where the gap is widest. A mapping drone with a good RGB camera is a modest purchase. A survey-grade lidar payload costs many times more, and it usually needs a heavier aircraft, which brings more complex operations and more demanding pilot requirements in some jurisdictions.
- Hardware: photogrammetry uses cameras many drones already carry; lidar requires a dedicated payload and aircraft.
- Field time: similar per acre, though lidar can fly higher and wider on some jobs, and it is less sensitive to light.
- Processing: photogrammetry is compute-heavy, which is why cloud processing is popular; lidar processing is lighter on compute but heavier on calibration and strip adjustment expertise.
- Skills: both need a trained crew; lidar adds GNSS and IMU quality control that many teams outsource at first.
For a firm starting a drone program, photogrammetry is usually the first step. It covers the majority of construction, mining, aggregate, agriculture and inspection work, and it produces deliverables that clients already understand. Lidar is added when the job mix includes vegetated terrain often enough to justify the investment.
When to choose photogrammetry
- Construction progress, site grading and earthworks on cleared ground.
- Stockpile inventory in yards and quarries.
- Roof, facade and asset inspection where the photo is the evidence.
- Agriculture mapping, including multispectral NDVI and NDRE, which are image-based by nature.
- Any project where the client wants an orthomosaic and a 3D model they can look at.
When to choose lidar
- Topographic surveys under forest canopy or dense brush.
- Utility and transmission line corridors with thin wires.
- Flood modelling and drainage studies where the true ground under vegetation matters.
- Low-texture surfaces where image matching is known to fail, such as snowfields.
Common misconceptions in the photogrammetry vs lidar debate
- Lidar is always more accurate. It is not. On open, textured ground with good control and a low flight, photogrammetry often matches or exceeds lidar in vertical accuracy, and it clearly wins on horizontal detail.
- Photogrammetry cannot produce a point cloud. It does. A dense photogrammetric point cloud is exported as LAS or LAZ just like a lidar cloud, and it carries true color for every point.
- Lidar needs no ground control. Lidar still benefits from control and always needs checkpoints. GNSS and IMU errors, boresight misalignment and datum mistakes show up as offsets that only independent checks reveal.
- More points means a better survey. Point density helps only when the points are accurate. A dense cloud with a vertical bias is still wrong everywhere.
Questions to ask before choosing a sensor
- What share of the jobs you fly each year have dense vegetation over the ground you need to measure?
- Does the client want an orthomosaic and a 3D model, or only a terrain surface and contours?
- What accuracy does the contract state, and how will it be verified with checkpoints?
- Who on the team will handle calibration, processing and quality control for each method?
Honest answers to these four questions usually settle the decision faster than any spec sheet.
Using both on the same project
Many survey teams combine the methods. A common pattern is lidar for the bare-earth terrain in wooded areas and photogrammetry for the orthomosaic and for the cleared parts of the site. Both datasets land in the same coordinate system, often defined by an EPSG code, and are tied to the same ground control. The point clouds are exchanged as LAS or LAZ, the imagery as GeoTIFF, and the terrain feeds contours in DXF or SHP for CAD and GIS.
The practical rule is simple. Decide on the deliverable first, then pick the sensor. If the deliverable is a picture of the site and a surface of what is visible, fly a camera. If the deliverable is the ground under something the camera cannot see through, fly lidar.