What Is an Orthomosaic? How Hundreds of Drone Photos Become One Accurate Map

If you've flown an autonomous grid mission and processed the imagery through DroneDeploy or Pix4D, you've produced an orthomosaic — you may just not have called it that. The finished map your client can zoom into, measure across, and drop into their project management platform is an orthomosaic.

Understanding what it is at a technical level makes you a better operator, a more credible consultant, and someone who can have an intelligent conversation with the surveyors, engineers, and project managers who use your deliverables every day.

The Problem Orthomosaics Solve

A single aerial photograph taken from a drone is not a map. It's a picture that looks like a map, but it has a fundamental flaw: it's a perspective projection, not an orthographic one. Objects closer to the camera appear larger. Tall features like buildings, trees, and stockpiles lean outward from the centre of the frame. The edges of the image are distorted relative to the centre. If you tried to measure distance between two points on a raw drone photo, you'd get the wrong answer.

In commercial applications, that distortion has real consequences. A roofing contractor using your imagery to estimate a repair needs accurate area calculations. A civil engineer using your data to track earthwork progress needs accurate volume measurements. A GC comparing this week's site to last week's needs images that line up precisely. Raw drone photos can't deliver any of that reliably.

An orthomosaic corrects for all of these distortions. Every pixel is repositioned to represent its true ground location. The result is a map you can measure on — not just look at.

How the Stitching Works

The process of building an orthomosaic starts in the air, long before any processing happens. It depends entirely on one thing: overlap.

Why Overlap Is Everything

When you fly a grid mission, your drone captures hundreds of individual images — each one a slightly different view of the same ground from a slightly different position. The key is that each image shares significant coverage area with its neighbours. This shared area is what the software uses to stitch everything together.

There are two types of overlap in a grid mission. Frontlap (also called forward overlap) is the percentage of coverage shared between consecutive images along a flight line — typically 75–85% for mapping work. Sidelap is the percentage shared between adjacent flight lines — typically 65–75%. Together these ensure that every point on the ground is photographed from at least three different positions, which is the minimum the software needs to reconstruct accurate geometry.

What looks like redundant photography from above is actually the raw material for everything that follows. The more overlap you fly, the more reference data the software has, and the more accurate and complete your final map.

Structure from Motion — How the Software Thinks

Once your images are uploaded to processing software like DroneDeploy or Pix4D, the engine runs a process called Structure from Motion (SfM) — the computational technique that turns a pile of flat photographs into geometric understanding of the real world.

SfM works by identifying thousands of matching feature points across multiple images — the corner of a roof, a painted line on pavement, a rock in a field — and calculating where in three-dimensional space each point must be located for all the camera positions to produce the images you captured. The result is a dense point cloud: millions of georeferenced 3D points representing the surveyed surface.

From that point cloud, the software generates several outputs. One is a Digital Surface Model (DSM) — a raster grid where every cell contains an elevation value, including the tops of buildings and trees. Another is a Digital Terrain Model (DTM) — the same thing but filtered to represent bare ground, with above-ground features removed. And the third — the one most operators deliver most often — is the orthomosaic itself.

The Orthorectification Step

Building the orthomosaic from your processed imagery involves a step called orthorectification. This is where the geometric correction actually happens.

The software takes each original photograph and uses the calculated camera position and the DSM elevation data to systematically reproject every pixel from its perspective-distorted position to its true horizontal ground position. Tall features that appeared to lean in the original photo are straightened. Edges that were distorted by lens geometry are corrected. Scale variations between centre and edge of frame are eliminated.

The corrected images are then seamlessly blended together — the seams between individual photos are smoothed out using colour-matching algorithms so the final mosaic looks like a single continuous image rather than a patchwork of photos.

The result is an image where every pixel represents a consistent, measurable ground distance. That property is called geometric fidelity — and it's what makes an orthomosaic fundamentally different from any regular aerial photograph.

Orthomosaic vs. Photogrammetry — Two Terms That Get Conflated

These two terms get used interchangeably in the drone industry, including by operators who should know better. They're related but not the same thing.

In practice this means: when a client asks for "photogrammetry," you need to clarify what deliverable they actually want. They might want an orthomosaic. They might want a 3D model. They might want elevation data for volumetric analysis. These are different outputs with different processing times, different software requirements, and different value to the client. Knowing the distinction lets you scope a job accurately and price it correctly.

What Orthomosaics Are Used For

The orthomosaic's defining property — geometric accuracy — makes it useful anywhere a client needs to measure, monitor, or document a site from above. That covers more commercial use cases than most operators initially realise.

The Numbers Behind a Quality Orthomosaic

Understanding the technical parameters that drive orthomosaic quality separates operators who produce reliably accurate results from operators who sometimes get lucky. These are the variables you control — at mission planning time and at the processing stage.

Ground Sampling Distance (GSD)

GSD is the distance on the ground represented by a single pixel in your final orthomosaic. A GSD of 2 cm/px means each pixel in the finished map corresponds to a 2cm × 2cm area on the ground. Lower GSD means higher resolution — more detail per unit of area.

GSD is determined by three factors: flight altitude (lower = smaller GSD = higher resolution), camera sensor size, and focal length. Flying lower produces better resolution but requires more images to cover the same area, more processing time, and more battery cycles. The right GSD depends entirely on the application:

• Roof inspection and damage documentation: 1–2 cm/px — you need to see individual shingles and flashing detail
• Construction progress monitoring: 2–4 cm/px — adequate for measuring work completed without excessive flight time
• Large-area land mapping: 4–8 cm/px — efficient coverage of large parcels where individual feature detail is less critical
• Survey-grade work with RTK: 1–2 cm/px — combined with ground control points or RTK positioning for centimeter accuracy

Overlap Requirements in Detail

The standard mapping overlap parameters — 75% frontlap, 70% sidelap — are a reliable starting point but they're not universal. Several conditions require more:

• Flat featureless surfaces (sandy lots, open water, uniform grassland) have few identifiable feature points for SfM to match. Increase frontlap to 85%+ and sidelap to 80%+ to compensate.
• Tall structures and significant terrain variation create occlusion — areas hidden from the camera angle by nearby objects. Additional oblique imagery captures the hidden faces.
• Forests and dense vegetation are problematic for bare-earth DTM generation. Ground points are sparse and elevation accuracy suffers. Higher overlap improves the odds of capturing ground-level features through gaps in the canopy.
• Windy conditions can cause the drone to deviate from its planned flight path and alter the actual overlap captured. Flying in lighter wind or compensating with higher planned overlap protects against this.

Ground Control Points vs. RTK

A drone's onboard GPS positions each image with consumer-grade accuracy — typically 1–3 metres. For many applications that's fine. For survey-grade work it isn't.

Ground Control Points (GCPs) are precisely surveyed markers placed on the ground before the flight. Their known coordinates are entered into the photogrammetry software and used to anchor the model to real-world coordinates, correcting the GPS error systematically. A well-distributed set of 5–8 GCPs can reduce positional error to 1–3 cm.

RTK (Real-Time Kinematic) GPS achieves similar accuracy without GCPs by continuously correcting the drone's GPS signal against a base station or network correction service. RTK drones like the DJI Matrice 4E and P4 RTK can achieve 1–2 cm positional accuracy at the image level, which propagates through to centimeter-accurate orthomosaics — without requiring a survey crew to pre-place targets.

For most commercial AEC work, RTK is the more practical option. GCPs remain valuable as a check shot verification method and in areas where RTK network coverage is unreliable. See our guide to best drones for construction mapping for a breakdown of RTK platforms.

Processing Pipeline

The photogrammetry processing pipeline follows a consistent sequence regardless of which software you use:

1. Alignment / Keypoint matching — the software identifies matching feature points across overlapping images and calculates the position and orientation of every camera at the moment of capture
2. Dense point cloud generation — the software densifies the sparse keypoint cloud into a high-density 3D point cloud representing the surveyed surface
3. Mesh generation (optional) — the point cloud is converted into a 3D polygon mesh for model outputs
4. DSM generation — the point cloud is rasterised into a gridded elevation model
5. Orthorectification and mosaic blending — original images are geometrically corrected against the DSM and stitched into the final orthomosaic

Processing time scales with image count, overlap percentage, and the resolution setting you select. A 300-image job at standard quality takes 30–60 minutes on a capable workstation. A 2,000-image high-overlap survey at maximum quality can take several hours. Cloud processing via DroneDeploy offloads this to their infrastructure — useful if your computer isn't purpose-built for photogrammetry workloads. See our DroneDeploy vs Pix4D comparison for a full breakdown of processing workflows and software tradeoffs.

Output Formats and Client Delivery

Orthomosaics are typically delivered as GeoTIFF files — a raster image format that embeds the coordinate reference system and georeferencing data directly in the file. A GeoTIFF can be opened in GIS platforms (QGIS, ArcGIS, AutoCAD Civil 3D), project management platforms (Procore, Autodesk Construction Cloud), and web viewers.

For clients who don't have GIS software, an interactive web map delivered through DroneDeploy's viewer or a similar platform is often more practical — they can zoom, measure, and annotate directly in a browser without installing anything. This is one area where DroneDeploy has a significant advantage over Pix4D for client-facing deliverables.