The Pixels-to-Points™ tool requires a series of images with 40% or more overlap (optimally at least 60% overlap) between subsequent images.
- For good 3D reconstruction, the area of interest must have texture. This means there should be distinct features or patterns in the image. Images of the sky and bright snow fields do not reconstruct well. It is recommended to remove images containing significant amounts of sky.
- Choose a collection day with low wind and low cloud cover. Low wind is especially important in forested areas, where motion in the trees can cause 3D reconstruction to fail. Shadows, and especially moving shadows, can also present issues with reconstruction.
- This tool relies on the motion of the camera, while the scene remains static. Objects that move during the time the images are taken, such as waves or moving vehicles, will generally not appear in the output data because they do not match properly.
- It is recommended to not set the camera on auto-focus. Images that have changing aperture and shutter speed can cause issues with matching during 3D reconstruction.
Ground Control Point Collection
Ground Controls are not required for the Pixels-to-Points tool, but they can be used to improve the positional accuracy of the point cloud. The Pixels-to-Points process can obtain relative accuracy using just the image metadata and algorithm, but the absolute position of the output may be less accurate. There are additional tools available to improve the absolute position during post processing, see Rectification and QC of Lidar Data
For more information on Ground Control see Ground Control Points with Pixels-to-Points™
- Obtain high accuracy location coordinates for identifiable features throughout the scene being collected. A common procedure for this is to place targets on the ground throughout the scene during the image collection process.
- A minimum of 3 ground control points are necessary, and should ideally be located on the ground in open areas spread throughout the scene.