Photogrammetry 101: A Beginner's Guide to 3D Modeling

What Is Photogrammetry?

Photogrammetry is the process of creating accurate 3D models and measurements from overlapping 2D photographs. The name comes from three Greek roots: photo (light), gramma (drawing), and metron (measurement).

Software finds the same features in many photos taken from different positions. It then uses triangulation to work out where each feature sits in 3D space, much like your two eyes judge depth.

The result can be a point cloud, a 3D mesh, a textured 3D model, or a scaled map. Engineers, surveyors, archaeologists, and inspection teams use these outputs to measure, document, and compare real-world objects and sites.

What Is 3D Photogrammetry?

3D photogrammetry, also called photogrammetric modeling, builds a full three-dimensional model of a subject. It is what most people mean when they talk about turning photos into a 3D model.

2D photogrammetry, by comparison, measures positions and distances on a flat plane, such as a scaled map or orthophoto. Many projects use both: a 3D model for inspection and a 2D map for planning.

How Does Photogrammetry Work?

Photogrammetry works by matching shared features across overlapping photos and calculating where the camera was for each shot. Once the camera positions are known, the software can triangulate millions of points on the subject's surface.

Most photogrammetry software follows the same core workflow, from image capture to a finished, scaled 3D model.

Step 1: Capture Overlapping Images

Photograph the subject from many positions so that each image overlaps the ones around it. Capture quality sets the ceiling for the final model, so the next section covers it in detail.

Step 2: Align the Images

The software detects distinctive features in each image, such as edges, corners, and patterns, and matches them across the dataset.

This stage, often called structure from motion (SfM), estimates where each photo was taken and produces a sparse point cloud of the subject.

Step 3: Build a Dense Point Cloud

With the camera positions solved, multi-view stereo (MVS) algorithms compare pixels across images to calculate depth over the whole subject. The result is a dense point cloud, where each point holds X, Y, and Z coordinates plus color.

Step 4: Generate the Mesh

The software connects neighboring points into a continuous surface made of small triangles, called a mesh. The mesh gives the model its solid shape and lets you view it from any angle.

Step 5: Apply Texture and Scale

Color from the original photos is projected onto the mesh to create a photorealistic texture. The model is then scaled with a known reference, such as a scale bar, surveyed control points, or a calibrated stereo camera, so measurements match the real world.

How to Take Photos for Photogrammetry

Software can only reconstruct what the camera records. Sharp, evenly lit, well-overlapped images produce clean models. Blur, gaps, and lighting changes lead to holes, noise, and alignment errors.

Plan the Capture

Start by defining the subject and the area you need to cover. Then decide on camera positions, distance from the subject, and the path you will follow.

Factor in the conditions that affect image quality, such as available light, weather, and, underwater, water clarity and current. A few minutes of planning can prevent a return trip to fill gaps.

How Much Image Overlap Do You Need?

Aim for 50 to 80% overlap between consecutive images. Overlap gives the software enough shared features to match photos and triangulate points accurately.

Plan for overlap in both directions: front to back along your path, and side to side between passes. Extra passes at different heights or angles add coverage and improve accuracy.

Common capture paths include spirals, wave patterns, and straight parallel lines. Choose the pattern that keeps a steady distance from the subject and covers the full area.

Cover Every Angle

A complete model needs views of every surface. For an object, that usually means a full 360-degree pass plus shots from several heights.

Useful perspectives include:

• Top-down view: Records upper surfaces and the overall layout.
• Front-facing view: Shows the main face of the subject in full.
• 360-degree orbit: Circles the subject so no side is missed.
• Close-up shots: Capture texture and fine detail.
• Diagonal views: Connect top and side surfaces so the software can link them.
• Varying heights: Add depth by shooting the subject from different elevations.

Close-ups record fine texture, while wider shots help the software connect the whole subject. Use both, and avoid large jumps in distance between consecutive images.

If something blocks part of the subject, take extra images from other angles to fill in the hidden areas.

How Many Photos Do You Need for Photogrammetry?

It depends on the subject's size, shape, and complexity. A small, simple object may need a few dozen images. A large or detailed structure can take hundreds or thousands.

More images are not always better. Consistent overlap and sharp focus matter more than raw image count.

Camera Settings for Photogrammetry

Use manual settings so every image looks the same. Shifts in exposure or color between shots make features harder to match.

• Shutter speed: Use a fast shutter speed to prevent motion blur.
• Aperture: Use a higher f-number for a deep depth of field, so more of the subject stays in focus.
• ISO: Keep ISO as low as the light allows to reduce noise.
• White balance: Set it manually and leave it fixed for the whole capture.
• Focus: Keep focus consistent and check sharpness as you go.

How Lighting Affects Model Quality

Even, consistent lighting reduces the shadows, glare, and color shifts that confuse photogrammetry software. Avoid harsh contrast and changing light during a capture.

Outdoors, overcast days give soft, even light. Indoors or underwater, diffusers and steady artificial lighting help keep exposure uniform from the first image to the last.

Choosing a Camera for Photogrammetry

You do not need specialized equipment to get started, but these features make a clear difference in results:

• Image resolution: More pixels give the software more detail to match.
• Sensor size: Larger sensors handle low light with less noise.
• Lens quality and focal length: Sharp lenses with low distortion improve precision.
• Manual controls: Fixed settings keep images consistent.
• Stabilization: Reduces blur when shooting handheld or from a moving platform.
• Interval or time-lapse shooting: Captures images at a steady rhythm for consistent overlap.

Calibrate the camera to correct lens distortion, or let the software estimate calibration from the images. Most photogrammetry software accepts common formats such as JPEG.

Subjects That Are Hard to Model

Photogrammetry performs well on textured, matte, and varied surfaces. Smooth, shiny, single-colored, transparent, or mirrored subjects give the software few reliable features to match.

For difficult surfaces, add temporary markers for texture, adjust lighting to cut glare, or pair photogrammetry with another sensor, such as sonar or a laser scanner.

Photogrammetry Software

What Photogrammetry Software Does

Photogrammetry software automates the processing workflow: aligning images, building the point cloud, meshing, and texturing. Most packages also include tools for scaling, measuring, cleaning up models, and exporting to other formats.

Widely used options include Agisoft Metashape, Pix4D, 3DF Zephyr, and RealityScan (formerly RealityCapture), along with open-source tools such as Meshroom and COLMAP.

Finished models are commonly exported as OBJ, PLY, or FBX meshes, or as point clouds for use in CAD, GIS, BIM, and inspection software.

How to Choose Photogrammetry Software

The right package depends on your subject, your output, and your hardware. Compare options on:

• Input support: Still images, video frames, and stereo image pairs.
• Scaling tools: Support for scale bars, control points, and calibrated cameras.
• Measurement and reporting: Built-in tools for distances, areas, volumes, and annotations.
• Export formats: Compatibility with the CAD, GIS, or inspection tools your team uses.
• Hardware requirements: GPU, memory, and storage needs for your typical dataset size.
• Licensing: Free, subscription, or perpetual licenses, and any limits on commercial use.

Hardware and Processing Time

Photogrammetry processing is computer-intensive. A capable GPU, plenty of RAM, and fast storage shorten processing times considerably.

Small datasets may finish in minutes. Large, high-resolution projects can take hours or even days, depending on image count and hardware.

How Accurate Is Photogrammetry?

Photogrammetry can produce measurement-grade results, which is why it is used in surveying and metrology. Accuracy depends on how the images are captured and how the model is scaled. The main factors are:

• Resolution and distance: Higher resolution and a closer standoff capture finer detail per pixel.
• Overlap and coverage: More shared features mean stronger triangulation.
• Camera calibration: Correcting lens distortion removes systematic error.
• Scale reference: Scale bars, surveyed control points, or a fixed stereo baseline tie the model to real-world units.
• Image sharpness: Blur and noise reduce the number of reliable matches.

Without a scale reference, a model has the correct shape but no real-world size. Adding one turns a 3D picture into a measurement tool.

Photogrammetry Applications and Examples

Photogrammetry is used wherever a team needs an accurate visual and dimensional record. Common applications include:

• Surveying and mapping: Precise maps and terrain models that simplify field measurement.
• Architecture and construction: As-built records, renovation planning, and BIM models.
• Cultural heritage and archaeology: Digital records of artifacts and sites that can be studied without handling them.
• Film, games, and visual effects: Realistic 3D assets captured from real objects and locations.
• Forensics: Detailed, measurable reconstructions of scenes for investigation.
• Infrastructure inspection and digital twins: Models of dams, bridges, tanks, ship hulls, and subsea assets that show change over time.
• Marine science: Coral reef mapping, habitat monitoring, and seabed documentation.

Underwater Photogrammetry

Underwater photogrammetry follows the same principles, but water makes every step harder. It is used for hull, dam, and subsea asset inspections, as well as underwater archaeology and reef monitoring.

Challenges of Underwater Photogrammetry

• Turbidity: Suspended particles reduce visibility and add noise.
• Light attenuation: Water absorbs light quickly, starting with reds, so color and brightness change with depth and distance.
• Backscatter: Light reflecting off particles creates bright specks that interfere with feature matching.
• Refraction: Water bends light and lengthens the effective focal length, so cameras should be calibrated for underwater use.
• Motion: Current and surge move the camera, causing blur and uneven overlap.
• Disturbed sediment: Thrust near the seabed can stir up debris and cloud the scene.

Using ROVs for Underwater Photogrammetry

Underwater ROVs give operators steady, repeatable control of the camera and can stay on task through long image runs. They carry high-resolution cameras and lights, hold position against current, and record position data that helps confirm coverage.

Deep Trekker's Mission Planner provides assisted navigation along a predefined route, so the ROV can follow a planned capture path while recording images from multiple perspectives.

Dead reckoning displays the ROV's track line on screen, making it easier to confirm coverage and return to a specific point on a structure.

For vehicle settings, white balance tips, and a field case study, see A Guide to Using Deep Trekker ROVs for Photogrammetry.

SPECTRA ROV: Purpose-Built for Underwater Photogrammetry

Everything covered above, including sharp imagery, calibration, steady capture, consistent lighting, and strong overlap, is what SPECTRA was engineered to deliver.

SPECTRA is a portable inspection and survey ROV that pairs a factory-calibrated stereo 4K camera with seven-thruster propulsion and real-time 3D perception. It turns the capture practices above into a repeatable field workflow.

Factory-Calibrated Stereo 4K Camera

The camera is the main reason SPECTRA produces reliable photogrammetric results. It is a fully integrated, rotating stereo 4K system that captures 12 MP still images and 4K video at up to 60 fps.

• Metric scale out of the box: The two lenses sit at a fixed, factory-calibrated baseline, so depth comes directly from each synchronized image pair. You can build a scaled 3D model without placing scale bars or reference markers.
• Calibration built in: Lens distortion correction is handled at the factory instead of in post-processing.
• Every angle with less repositioning: The camera head rotates up to 180 degrees, so you can image above, below, and around a structure while the vehicle holds position.
• Integrated, not bolted on: Stereo vision is part of the vehicle design rather than an add-on skid, so handling stays balanced during capture.

Thrust and Stability That Protect Image Quality

Blur, drift, and stirred-up debris degrade a dataset. SPECTRA's seven independently controlled brushless thrusters, four vectored and three vertical, produce 40 kg (88 lbs) of thrust in a symmetrical layout.

That layout holds station in currents up to 2.3 knots from any direction, reaches forward speeds over 3.5 knots, and gives strong lateral control for holding a line along a structure.

• Minimal drift: Precise low-speed control keeps framing steady, which means sharper images and less motion blur.
• Consistent overlap: A stable heading and standoff distance make it easier to hold 50 to 80% overlap and repeat a spiral, wave, or straight-line path.
• Cleaner water: Steady station-keeping reduces the need for aggressive thrust near the seabed, so less debris clouds the scene.

Lighting That Tracks the Camera

SPECTRA carries a 300,000 lumen LED array that follows the camera's orientation, keeping light aligned with the lens wherever it points.

Paired with a large-format, high-sensitivity sensor, it delivers clear, evenly lit imagery for photogrammetric processing, even in dark water down to the vehicle's 1,000 m (3,280 ft) depth rating.

Real-Time Coverage Checks With 3D Sonar SLAM

With 3D sonar equipped, SPECTRA builds a live point cloud of the work area through Sonar Simultaneous Localization and Mapping (SLAM), even in high turbidity.

For photogrammetry, that means coverage confidence. You can confirm every face of an asset has been captured, and spot gaps, before leaving the site. All SLAM processing runs onboard, so no topside computer is needed.

The SPECTRA Photogrammetry Workflow

A typical capture-to-model workflow with SPECTRA looks like this:

1. Plan the route around the target, setting standoff distance and the path pattern needed for full coverage.
2. Capture the dataset by starting a recording in the vehicle interface. The stereo camera collects synchronized image pairs while the lights track the camera head.
3. Verify coverage live using the 3D Sonar SLAM point cloud before demobilizing.
4. Export post-mission, including stereo datasets and, if equipped, SLAM point clouds and navigation tracks.
5. Process into a model by importing the stereo image pairs into standard photogrammetry software for a photorealistic, metrically scaled 3D reconstruction.

The output supports site mapping, asset visualization, defect measurement, and year-on-year change detection. A single inspection becomes a measurable, repeatable record.

Photogrammetry FAQ

What Are the Steps in Photogrammetry?

The core steps are image capture, image alignment, dense point cloud generation, meshing, and texturing. Scaling the model to real-world units is usually the last step before export.

Is Photogrammetry the Same as 3D Scanning?

Photogrammetry is one type of 3D scanning. It builds geometry from ordinary photographs, while laser and structured light scanners measure distance directly.

Photogrammetry captures realistic color and texture and scales to large sites with modest equipment, which is why many teams combine it with other scanning methods.

Can You Use Video for Photogrammetry?

Yes. Many workflows extract still frames from video. Dedicated still images usually give sharper results, since video frames are more prone to motion blur and compression. Interval or timed snapshots are a good middle ground.

Do You Need a Drone for Photogrammetry?

No. Any camera can be used. Aerial drones suit land surveys and large sites, handheld cameras suit objects and rooms, and underwater ROVs handle submerged structures.

How Do You Learn Photogrammetry?

Start with a small, textured object you can walk around. Photograph it in even light with plenty of overlap, process it in a free or trial software package, and review where the model fails.

Each hole or blurry patch points to a capture fix, such as more overlap, steadier framing, or better lighting. Repeat the process and the improvements show up quickly.

Work With Deep Trekker

Deep Trekker builds portable, easy-to-deploy underwater ROVs with 4K cameras, powerful LED lighting, and stable control for photogrammetry in hazardous and hard-to-reach environments.

We design our vehicles around customer feedback, and our engineering and support teams are a call or email away when you need guidance on a project.

Contact our team to talk through your photogrammetry application, or request a customized quote for your own vehicle.

Learn More About Underwater Photogrammetry and 3D Modeling:

• A Guide to Using Deep Trekker ROVs for Photogrammetry
• 3D Modeling Using Photogrammetry – BRIDGE Technology
• Improving Underwater Asset Management with ROVs
• Utilizing ROVs and Photogrammetry for Improved Inspections of Nuclear Plants
• How Deep Trekker’s Enhanced 4K Camera Elevates Underwater Inspections