Photogrammetry is a technique for obtaining information about the shape and position of an object by measuring and interpreting multiple photos of it taken from different positions. The U.S. Geological Survey (USGS) defines it as "the science of obtaining reliable information by measuring and interpreting photographs" (source: USGS "Topographic Mapping", as of July 2026). In this article, we walk through how it works, the gear you need, how it relates to 3D scanning with a smartphone, and how it differs from AI that generates 3D from images.

Note: Figmee currently lets you upload drawings and illustrations to generate and download figurine-style images and 3D model data (GLB / 3MF). Sales of physical 3D-printed items are in preparation (Coming Soon).

How It Works: Why Photos Can Reveal Shape

The underlying idea is the same one that lets humans perceive depth with two eyes. When you look at the same object from two positions a little apart, it looks slightly different from each. How large that difference is depends on the distance to the object: nearby things shift a lot, distant things shift only a little. Work this relationship backward and you can calculate distance from photos.

The USGS explains that viewing pairs of overlapping aerial photos stereoscopically produces a three-dimensional image, and that topographic maps have long been made this way. The method is said to have replaced traditional plane-table surveying in the 1940s, dramatically changing the scope and accuracy of mapmaking (source: USGS "Topographic Mapping", as of July 2026).

Modern photogrammetry automates this idea on a computer. Processing broadly happens in three stages.

  1. Detecting and matching feature points: The software finds landmark points such as patterns and corners in each photo, then works out correspondences like "this point in this photo is the same spot as that point in that photo."
  2. Estimating camera positions and generating a point cloud: From those correspondences, it calculates backward to figure out the position and angle each photo was taken from, while also determining the 3D positions of points on the object's surface. The result of this stage is a point cloud.
  3. Building surfaces and applying textures: The points are connected into surfaces (a mesh), and colors taken from the original photos are applied to the surface.

The first stage, matching, decides whether the whole approach succeeds. If the software cannot find the same spot across photos, none of the calculations that follow can work.

The Gear You Need

It may come as a surprise, but you do not need special measuring equipment to try photogrammetry. You need just three things.

What you needMinimum requirementNice to have
CameraEven a smartphone camera worksA bright lens and a tripod to keep it steady
Shooting environmentBright, even lightA turntable for the object and a plain background
Processing softwareSoftware that supports photogrammetryA PC powerful enough to handle the processing

The main cost is not gear but time spent shooting and processing. You need to circle the object, changing the angle as you go, and take a large number of shots so that neighboring photos overlap enough. The calculations afterward also take time.

Here are the key points to keep in mind when shooting.

  1. Circle the object, changing the angle a little at a time (guide: 15 to 60 minutes) — Move in small steps so each photo overlaps heavily with the one next to it.
  2. Keep the lighting the same (guide: throughout the shoot) — If the way light hits the object changes mid-shoot, the same surface may be treated as something different.
  3. Do not move the object (guide: throughout the shoot) — If the position of the object relative to the background changes, the calculations break down. If you use a turntable, you need to rotate the background along with it.
  4. Keep everything in focus (guide: throughout the shoot) — Blurry photos make feature point detection fail.

How It Relates to 3D Scanning with a Smartphone

What people call "3D scanning with a phone" actually covers two different methods.

The method that reconstructs from photos is photogrammetry itself. Because it uses multiple photos taken with the camera, it works even on devices without special sensors.

The method that measures distance with light uses depth sensors such as LiDAR. NOAA describes lidar as short for "Light Detection and Ranging," a remote sensing method that uses pulsed lasers to measure distances. It explains that the output is "point clouds" made of points with three-dimensional coordinates, and that the system consists of a laser, a scanner, and a GPS receiver (source: NOAA "What is lidar?", as of July 2026).

In other words, photogrammetry is one of the methods within smartphone 3D scanning. Which apps you can use depends on whether your device has those sensors, so check which method your phone supports before you try, and you will save yourself some trouble. We compare the two approaches in 3D Scanning vs. AI Generation Compared.

How It Differs from AI Image-to-3D Generation

In recent years, AI that builds a 3D model from a single image (image-to-3D) has also become widely used. It is easy to confuse with photogrammetry, but the two rest on completely different premises.

Point of comparisonPhotogrammetryAI generation (image-to-3D)
What it doesMeasures photos to reconstructEstimates and creates from learned knowledge
SubjectsOnly real objects and spacesAlso drawings and characters that do not exist
Photos neededMany overlapping photosCan work from just one
Parts not capturedRemain as missing dataFilled in by estimation
Dimensional accuracyHigh when conditions are rightNot guaranteed
Main usesRecording, measurement, replicationCreative work, prototyping, checking ideas

The decisive difference is how each handles the parts that were not captured. Photogrammetry leaves the spots it could not photograph as holes, treating them as "not there." Since it is a form of measurement, it cannot fill in places it never measured. AI generation, on the other hand, fills in the unseen back by estimation. That filling-in is exactly what lets a single image become 3D, and at the same time there is no guarantee those parts match the original subject.

We cover how the back gets filled in when you start from a single drawing, and how to check it, in Image to 3D: What Happens to the Back Side When AI Works from a Single Image?.

So the division of roles is this: if you want to accurately record a real object, use photogrammetry; if you want to turn a character that does not exist into 3D, use AI generation. We explain how the AI side works in detail in How Image-to-3D AI Works.

Strengths and Weaknesses

Photogrammetry has clear strengths and weaknesses that come straight from its principles.

What it handles well

  • Subjects with patterns or texture on the surface, where the same spot is easy to find across photos.
  • Objects with a matte finish, such as stone, wood, fabric, and brick.
  • Subjects that do not move much, such as outdoor buildings and terrain.

What it struggles with

  • Transparent or translucent things. Light passes through, so the surface position cannot be pinned down.
  • Mirror-like or highly glossy surfaces. Reflections change with each camera position, so points cannot be matched as the same.
  • Plain, smooth surfaces. No feature points can be found, so there is nothing to match on.
  • Fine hair and fibers. Outlines shift from photo to photo, so the shape never settles.
  • Things that move during the shoot. People and pets come out distorted unless they can hold the same pose.

You can use this list as-is as a "checklist before you shoot." For subjects that fall under the weak conditions, there are limits to what clever shooting can fix, so considering another method may be quicker.

Common Failures and Their Causes

These are the four things that most often trip people up on their first try. Each can be explained by the underlying principles, so once you know the cause, you can deal with it.

Part of the shape is missing. Most likely, there were not enough photos of that side. If you circle the object only horizontally without changing height, you end up short on information about the top and bottom. Reshooting with added high and low angles improves the result.

Surfaces come out bumpy and wavy. This happens when feature point matching is unstable. It is a common symptom on plain, smooth surfaces and on glossy ones. Countermeasures include softening the lighting and laying a patterned cloth as the background.

The object splits into two separate objects. This happens when the object or background moved partway through the shoot. Shooting on a turntable changes the relationship between object and background, so either rotate the background along with it, or switch to keeping the object fixed and moving the camera instead.

Processing never finishes or freezes. Your processing environment does not have enough power for the number and resolution of the photos. First run it once with fewer photos and lower resolution settings, then raise the settings step by step to isolate the cause.

Where It Is Used

Photogrammetry is not a new technology. It has long been used in mapmaking. As mentioned above, the USGS has used it to create topographic maps since the 1940s. Today, thanks to faster computers, it has spread into areas such as the following.

  • Surveying and mapping: Recording terrain, calculating earth volumes, and assessing conditions after disasters.
  • Recording cultural heritage: Capturing sculptures and ruins in 3D without touching them.
  • Game and film production: Scanning real rocks, walls, and the like to use as textured background assets.
  • Architecture and real estate: Recording existing conditions and checking them against drawings.

What all of these share is the goal of "preserving something that actually exists, exactly as it is." Whether your goal matches this is what decides which method to choose.

When this technique is used in creative fields, it is usually in the context of "making assets." A typical use is capturing the texture of real rock faces or wood and using it in the backgrounds of fictional worlds. Rather than creating the characters themselves, it works as a way to borrow textures from reality. Here too, its role is distinct from AI generation, whose purpose is turning characters into 3D.

Sorting Out the Terms

This is a field with many words used in similar situations, which makes it easy to get confused. Here is how they relate, one line each.

  • Photogrammetry (photo measurement): A technique for obtaining 3D information by measuring and interpreting multiple photos.
  • 3D scanning: An umbrella term for methods that turn real objects into 3D data. Photogrammetry is one part of it.
  • LiDAR: A method that measures distance with light. The other major type of 3D scanning.
  • Point cloud: A collection of points with 3D coordinates. Intermediate data produced partway through processing.
  • Mesh: Points connected into surfaces. The common format for representing 3D models.
  • Texture: Color and pattern information applied to the surface of a 3D shape.

For the bigger picture of the terminology, also see the 3D Modeling Glossary.

FAQ

Q. Are photogrammetry and 3D scanning the same thing? A. Not exactly. 3D scanning is an umbrella term for methods that turn real objects into 3D data, and photogrammetry is the method within it that "reconstructs from photos." The other major type measures distance with light, as LiDAR does. The goals are similar, but the gear required and the subjects each struggles with differ.

Q. How many photos do I need? A. It depends on the size and complexity of the subject, so there is no single answer. The common requirement is that "neighboring photos overlap enough." If you skip large angles between shots, matching fails, so changing position a little at a time matters more than any target number of photos.

Q. Can I turn an illustration or a character drawing into 3D with photogrammetry? A. No. Photogrammetry works by photographing and measuring a real object from multiple angles, so it cannot be applied to subjects that do not exist. If you want to turn a drawn character into 3D, your options are AI generation of 3D from images or modeling in 3D software.

Q. How much difference does a smartphone versus a DSLR make for shooting? A. How you shoot affects the result more than your gear does. Photos in focus, consistent brightness, and enough overlap between neighboring shots: if you meet these three conditions, even a smartphone camera can capture the shape. A DSLR has the edge in dark environments or when you want to resolve fine detail. The realistic approach is to start by getting the conditions right with the gear you already have.

Q. Can I 3D print the resulting data as-is? A. Often that is difficult. Parts that could not be measured remain as holes, and fine noise may be present, so the data needs to be cleaned up into a closed shape before printing. This step is needed regardless of method, and the same goes for AI-generated data.

Summary

Photogrammetry is a technique for reconstructing the shape of a subject by measuring and interpreting multiple photos, and it is also called photo measurement. You can get started without special measuring equipment, but it requires many overlapping photos and processing time, and it struggles with transparent, glossy, and plain subjects. In smartphone 3D scanning, this photo-based method coexists with methods that use sensors such as LiDAR. It differs fundamentally from AI image-to-3D generation in whether the parts not captured are "left as missing" or "filled in by estimation." If your goal is to record a real object, choose photogrammetry; if your goal is to turn a character that does not exist into 3D, choose AI generation.