You’ve probably heard that drones can be used to create 2D maps and even 3D models of structures, job sites, and buildings. But it’s not always easy to understand how a drone is able to do this — or what hardware and software you need to get started.
Photogrammetry is the process of taking two-dimensional photographs and processing them into three-dimensional, digital models of the real world. There’s no doubt that photogrammetry can be a complicated topic, but my goal is to make it as simple as possible.
Here’s a high-level overview of some of the basics of photogrammetry.
What hardware and software do I need to get started with photogrammetry?
You may think you need specialized sensors or expensive equipment like LiDAR to capture 3D data with your drone. But really, you just need three things to get started:
- A good camera drone, such as a DJI Mavic 2 Pro or Parrot ANAFI 4K
- A drone ground control station that can run survey flights, such as Skyward’s InFlight mobile app
- Photogrammetry processing software, such as Skyward Mapping & Modeling powered by Pix4D
Photogrammetry starts with photos from your drone. These images usually need to be “georeferenced,” which simply means that your drone tags each picture it takes with location data based on the GPS position of the drone. Your drone also includes other types of data, such as the altitude of the drone. Nearly all drones do this by default.
Those photos, and the data that goes with them, are the main ingredients in photogrammetry. To map an area, you just need a whole bunch of drone photos that overlap each other.
How do I capture good data for a photogrammetry model?
The amount of photos you need to collect, and the quality of each photo, depends on the level of detail you need from your model. Of course, the better the pictures you feed the engine, the better the model should turn out. So things like the camera’s sensor size, resolution, and calibration can affect your model’s quality quite a bit.
The best way to capture the photos you need is to use an automated survey flight. In a survey, the drone flies back and forth in a grid pattern over a set area. It captures images that overlap on all sides. Pix4D recommends that the photos you collect have a minimum of 75% vertical overlap and 60% horizontal overlap. For higher quality models, more overlap is better — that gives the processing engine extra data to work with.
When you want to create a 3D model from your images, you’ll also want to fly an orbit around the perimeter of the object you’re modeling. This helps the drone get a better look at the sides of the object. For better results, fly multiple orbits at different altitudes to capture even more angles of the object.
How does a photogrammetry processing engine work?
The next step is to process your drone photos using a photogrammetry engine like Pix4D. The engine starts by arranging the photos according to the location data tagged in each image. Then the engine analyzes each picture and compares it to its neighbors, identifying where those images overlap. It compares adjacent pictures to figure out which pixels match up, and which ones are different.
When the processing engine is creating a 2D map, which is called an orthomosaic, the engine stitches the images together and blends the overlapping part into one big, flat image of the area you’re mapping.
When the processing engine is creating a 3D model, it uses the overlapping portions of each image to estimate the third dimension: height and depth. With a good dataset that’s been collected properly, the engine can make these calculations based on many overlapping points from a wide variety of angles.
What kinds of data deliverables does a photogrammetry engine produce?
As mentioned above, photogrammetry engines can produce orthomosaics, which are 2D maps of the surveyed area. Since an orthomosaic does not need to calculate 3D points, it can often be produced very quickly. Companies often use orthomosaics to identify problems such as road damage or vegetation overgrowth, or to get an overview of large areas.
When a photogrammetry engine produces a 3D model, the output usually has two parts. First, there’s a point cloud consisting of all the points the processing engine has calculated from the photos. The point cloud forms the frame of the 3D model.
Secondly, there’s a photorealistic mesh. This is probably what you think of as a 3D model. Essentially, the engine stitches the photos together and stretches them out on top of the point cloud frame. With good data inputs and a reliable processing engine, it should produce an accurate digital representation of the area or structure being surveyed.
A model with a good detail and resolution can be measured with high accuracy when compared to the real-world object. You can estimate measurements like length, area, elevation, and volume. And if you build a map or model of the same area multiple times, you can compare changes over time.
How can I learn more about drone photogrammetry?
I’ve explained some of the basics of photogrammetry at the most fundamental, 101 level. If you want to learn more, check out Skyward’s Photogrammetry Training course. This is a day-long workshop for enterprise drone teams delivered by a Skyward expert trainer. It is designed to help drone programs go from no experience to producing high-value deliverables in a matter of weeks.
In it, we dive deeper into how photogrammetry works and cover topics such as:
- Ground sampling distance (GSD) and image quality
- Relative accuracy vs absolute accuracy
- Georectification: ground control points (GCPs) vs real-time kinematic (RTK) vs post-processed kinematic (PPK)
- Photogrammetry vs LiDAR
- Image capture techniques
- Skyward InFlight mobile app & automated flight modes
- Photogrammetric processing engines
If you’re interested in the Photogrammetry Training workshop, visit Skyward’s Pricing page for more details.
Where can I find photogrammetry software for my drone program?
Skyward can equip you with the photogrammetry solution that’s right for your drone team. Whether you need advanced, fine-tuned processing or you’re looking into 3D modeling for the first time, we can help you find what you need.
For those just getting started with photogrammetry, Skyward offers Mapping & Modeling, powered by Pix4D. This photogrammetry solution is integrated into the Skyward platform and uses the power of Pix4D, a market leader in photogrammetry software technology. The processing engine is in the cloud, simplifying much of the complexity of photogrammetry. You can process data from nearly any drone, analyze maps and models, and export for 3rd-party apps right from the Skyward platform. And it’s all part of a single Skyward subscription.
For companies who need deeper insights into their drone data, or those that require rapid mapping solutions, Skyward is also a premier reseller of Pix4D desktop and cloud applications. That means we can help you find the application that meets your data processing needs. This includes PIX4Dmapper for survey-grade photogrammetry and PIX4Dinspect for streamlined industrial inspections.
Whether you’re looking for an integrated photogrammetry solution or a third-party app, Skyward can help equip your drone program with what you need. Check out Skyward’s Mapping & Modeling page for more information on our integrations and partnerships.