Here is a question we get all the time:
How accurate is the 3D model I made using Elios 2 visual data?
The answer to this question is a little complicated, because the accuracy of your 3D model depends on the accuracy of the data you used to build it.
However, there is a general answer that can be applied to any 3D model or dataset:
The theoretical maximum accuracy of a 3D model created using visual data collected by the Elios 2 is three times the Ground Sampling Distance (i.e., the size of pixels as they appear on the surface of the model).
Since the Elios 2 has a 4K camera, you can achieve a maximum accuracy for Ground Sampling Distance (GSD) of 0.25mm/pixel. Therefore, the resulting model created using visual data with the highest accuracy possible collected by the Elios 2 would be an accuracy of 1mm (theoretically it would be .75mm, but we’re rounding up to provide a whole number that will be easier to remember).
This video shows how this level of accuracy can be achieved when using one or more reference dimensions close to the measured objects to scale the model:
Here are some key features of the model featured in the video:
- This video showcases an experiment performed in ideal conditions (i.e., no backlight is present; images were taken on a non-reflective surface; the surface has lots of texture).
- We only modeled a small area (about 0.5 m2) in order to measure small objects (<5cm).
- Two scale constraints and one orientation constraint were created across this area.
- The test measurements were made very close to the scale constraints.
- The model was not referenced in any specific coordinate system, so we considered the relative accuracy of the resulting model.
Key Considerations Regarding Accuracy in a 3D Model Made with Elios 2 Visual Data
While the video shown above demonstrates that you can achieve up to 1mm accuracy for a 3D model with the Elios 2, this level of accuracy is probably not realistic—or even desirable—for every project.
Your desired accuracy should always be considered in the planning stage of your project, and it should be determined based on what exactly you’re planning to do with your 3D model.
If you need to be able to drill down and know exactly how close one part of an object is to another to a great degree of accuracy then you’ll want to plan accordingly, fly very close to the object, and get the highest quality data possible.
But if you just need to have a general sense of where things on a surface are relative to each other, you don’t need to spend as much time planning or worry about getting as close as possible to the surface.
[Related reading: Why use photogrammetry to build 3D models of indoor spaces?]
Either way, the important point is that the quality and accuracy of the 3D model you create is entirely dependent on the quality and accuracy of the data you collect and use to make that model.
Here are some key considerations for ensuring accuracy in your 3D model:
- Larger areas require more reference points (called scale constraints in Pix4D). The larger the area you want to map, the more reference points need to be measured regularly along the flight path. Also, keep in mind that larger areas also require larger reference dimensions.
- Distance from the object matters. For accuracy, the closer you can get to the object the better (more information on this topic is covered in the next section on Ground Sampling Distance).
- Lighting is important. Dust and darkness (i.e., poor lighting) will affect the quality of the images you collect, and therefore the quality of the 3D model you create.
- Calibration is important. You want your drone flying as smoothly as possible so that your images are as clear as possible, with little to no blurring.
- Reflective surfaces are harder to map. Reflective surfaces offer fewer features, and will negatively affect image matching when it comes to making your 3D model.
Larger areas require more reference points
The Importance of Ground Sampling Distance (GSD) Resolution
Ground sampling distance (GSD) is the distance between the centers of two adjacent pixels measured on the observed object.
GSD is a type of camera resolution that is typically measured in mm/px. A GSD of 1mm/px means that one pixel on the image represents 1 mm in the real world.
Unlike pixel resolution, GSD resolution depends on the distance between the camera and the subject. This means that the GSD improves (i.e., its values get smaller) when the camera comes closer to an object.
A smaller GSD means that the object will appear bigger, and that smaller details will be more visible in the image.
How GSD Resolution Impacts the Accuracy of Your 3D Model
Photogrammetry is all scale related.
This means that the closer you can get to the object or area you are mapping—that is, the better your GSD resolution is—the more accurate your map will be.
Due to its unique cage design, the Elios 2 can fly very close to objects, which can make it a useful tool for collecting high-quality data that will produce high-quality 3D models. For example, in the video above the Elios 2 is flying just 20 cm (7.8 inches) away from the surface.
But again, the GSD resolution you choose should be determined by the needs of your project.
The tradeoff will always be between accuracy and speed. If you need a high level of accuracy in your 3D model, you should fly as close as possible to the surface or object and take as much time as you need to do so. On the other hand, if you don’t need a high level of accuracy in your 3D model then you can fly farther away, getting more coverage in a shorter period of time.
Did you know?
The Elios 2 can capture images with a GSD of 0.18 mm/px (0.007 in/px) when capturing data at a 30 cm (11.8 in) distance.
The three key types of resolution that matter for your work are pixel resolution, GSD resolution, and spatial resolution—learn more about how the Elios 2’s camera stacks up for all three.
GSD Fisheye Lens Considerations
When using a fisheye lens the GSD also depends on the position of the object within the image, not just the distance from the camera to the object (since parts of the lens are farther or closer to the object).
Objects in the center of the image will have a smaller GSD (i.e., they appear bigger, since that part of the lens is closer to the object) and objects in the corners will have a higher GSD (i.e., they appear smaller, since that part of the lens is farther away from the object).
Use Cases and Accuracy
We’ve already talked about the importance of considering the accuracy you want to obtain in your 3D model as you plan your mission.
But let’s take a step back and consider use cases and accuracy in general.
If your goal is to create a land survey, the level of accuracy you need will be much lower than the accuracy you’ll need if you’re conducting an indoor inspection, which might require you to understand how far away a weld or bolt is from an entryway, or from another weld or bolt.
To make this more concrete, if you’re flying the Elios 2 in the refractory lining of a 500mm duct in order to identify the location of a defect, you’ll want to know where that defect is with enough accuracy so that a person can enter the confined space nearby the place where work is needed.
This is all to say that, when considering accuracy for 3D models, the use case should be an essential part of your planning.
[Case study: Elios 2 tested for indoor stockpile volumetry, produces 3D maps accurate to within 1 centimeter]
Want to Learn More about Accuracy and Photogrammetry with the Elios 2?
Here are a list of resources to help you learn more about photogrammetry with the Elios 2:
- [Webinar] Indoor 3D Modeling Use Cases: Photogrammetry in Action
- [Webinar] Indoor 3D Modeling: Applications & Implications
- [Article] Why use photogrammetry to build 3D models of indoor spaces
- [Article] Building 3D models with Elios 2: How to acquire appropriate data for photogrammetry
- [Article] Building 3D models with Elios 2: Processing data with a photogrammetry software