‘Are you getting the accuracy you expect with Pix4Dmapper?’ 1
1 https://pix4d.com/getting-expected-accuracy-pix4dmapper/, last accessed 2/27/2018.
Project Summary: Original flight with a Phantom 4 Pro with 95% frontlap and 90% endlap and a 2.9cm ground sampling distance (GSD). The GSD is the distance between two consecutive pixel centers measured on the ground. For example, a GSD of 2.9 means that one pixel in the image represents linearly 2.9cm on the ground (2.9*2.9=8.41cm2).
First they generated an orthophoto with all of the data, then another with every third image, and a third ortho with every sixth image. Horizontal accuracy was measured by comparing the GPS locations of 16 ground control points (approximately 95%, 80%, and 73.5% image overlap).
In traditional photogrammetry, vertical accuracy is relative to the base-height ratio (this holds true for GPS accuracy as well). However, when processing UAV imagery, the software requires a high amount of overlap (~75% to 90%) when creating orthophotos and terrain models.
The “Skip 0” scenario gave the best results, though all yielded a horizontal accuracy of less than 1 pixel and a vertical accuracy of less than 3 pixels. Processing times ranged from two hours to ten minutes.
Figuring out the best image overlap scheme for your projects might take a little experimentation. A good starting point, and the solution I use most of the time, is 80% sidelap and 80% endlap. In more rugged or complex terrain, you might consider increasing your overlap (which will increase your flying time). In less-complex terrains, you might get by with decreasing your overlap and decreasing your flying time.
Planning a data acquisition mission…
The goal here is to determine
- what measurements you will make at the field site,
- how you will make them, and
- where you will make them
Do this before you get to the site so you will NOT need to revisit/refly an area because you forgot something.
- You need a high amount of image overlap (recall the “Are you getting the accuracy you expect with Pix4Dmapper Pro?” post we looked at Wednesday).
- Recommendations are a minimum of 70%. However, at this level, at certain sites, I have found this tolerance to be too low which results in holes in the resulting orthophoto and terrain model. Areas of steep terrain require more overlap than areas with little relief.
- The image processing software (Pix4D, Agisoft Photoscan, OpenDroneMapper, etc) automatically locates 1000’s of common points on adjacent photos in these overlapping areas. These points on one photo are referred to as keypoints, and when these locations are found on two or more photos, they are called matched keypoints.
- The software outputs a 3D point for each matched keypoint (X,Y, and Z values). This is the sparse point cloud.
- Recommendations by Pix4D for several scenarios are listed below:
- General case: For projects that do not include forests, snow, lakes, agricultural fields and/or other terrain that is difficult to reconstruct.
- Forest and dense vegetation: For project with areas covered by forest or dense vegetation.
- Flat terrain with agriculture fields: For flat terrain with homogeneous visual content such as agriculture fields.
- Building reconstruction: For 3D modelling of buildings.
- Special cases: For snow, sand, and water surfaces (oceans, lakes, rivers, etc).
- Corridor mapping: For projects with linear area of interest (roads, rivers, etc).
- Multiple flights: For projects with images taken using multiple flights.
- City reconstruction (visible facades): For 3D modelling of urban areas.
- 3D interior reconstruction: For 3D modelling of the interior of buildings.
- Mixed reconstruction: For combined datasets (interior/exterior and/or aerial/terrestrial and/or nadir/oblique).
- Large Vertical Objects reconstruction: For 3D modelling of objects like power towers, wind turbines, etc.
- Tunnel reconstruction: For 3D modelling of a tunnel.
- Small objects: For 3D modelling of small objects.
Image rate (photo interval)
Ground control (GCPs)