PhotoMesh provides a wide array of tools and features for evaluating and improving your aerotriangulation (AT).
- Project images on terrain - Once you load all the data into PhotoMesh, project photos from each of the different collections on to the terrain, and check if they fit with the terrain and each other.
- Sample ground control points to check the accuracy of the external orientation - In the Control Point Editor, if the white crosshairs (projected location of control point) is less than 3 pixels from the red crosshairs (user sample for position of control point), this indicates that the imported orientation information is relatively accurate. Otherwise, recheck the orientation information and other collection settings such as coordinate system, vertical datum, focal length, and principle point.
- AT report - The main indicator of the AT quality is the mean median error. The mean median error is the average of the median reprojection errors for all photos in the collection. Generally, a value lower than 0.5 pixels indicates that the AT's quality is good. It is important to check the mean median error of each of the collections.
- Camera Parameters - Compare the original focal length and principal point values to the calculated ones. A large difference between the two indicates that there are inaccuracies in the input data, either with the data elevation and coordinate system or the input focal length value.
- Tie Points and Check Points - Good ground control points (GCPs) from a surveyor or other verified source are important in validating AT accuracy. Ground control points associate real-world X, Y, Z coordinates obtained through surveying or sampling methods with corresponding locations (in pixels) in photos. Make sure you have both a good GCP scattering, i.e. that at least 5-6 ground control points have been placed in all the corners and center of each AT tile, as well as checkpoints (used only to sample the AT accuracy in the result, but not to improve the accuracy of the model’s position). When the AT is complete, check the GCP report and editor, investigate the photos whose control points have the highest errors, and resample if necessary. After fixing the GCPs, rerun the bundle stage of the AT. Below is an example of a good GCP report: