I did not understand your statement about the doming error being consistent between the two models. There is no first-order doming error in the final model. In fact there is no ground elevation information at all in this model. All ground is at 0 elevation across the entire site. There are still second-order errors due to varations within a grid-cell of the DTM (16’x20’) but these are small, amounting to less than 0.1’ across a cell. Could you elaborate on your statement?
Mind you I am all for using near zero-cost tricks to eliminate ugly doming elevation errors. Its just that in some cases where elevation measurements need to be made relative to the local ground level, their elimination appears to provides no benefit.
Here is an elevation map showing about a 7’ doming elevation error across an 80-acres site:
and here is the elevation map of the final model shown over a 1’ range (+/- 6 inches). In both cases the black dots are the produce sticking up from the ground.
The ground-level in the final model is quite flat.
SolarBarn: I have been assuming that you produced the bare-earth DTM from the “bare-earth” point cloud extracted from the overall point cloud. Where do you get the bare-earth model? If you are determining vegetation height, you should be comparing the vegetation model to some base model. Please explain.
Your assumption is close but there are some differences. Here are the steps I used to create the bare earth model and the final 3D model:
(1) A 3D mesh model is generated from the point cloud (2) This model is gridded and within each grid the ground points are identified. (3) The elevation of the ground points within each grid is averaged and this forms the DTM. (4) A 3D surface is fitted to the DTM which becomes the bare-earth model of the site. (4) For each vertex in the 3D mesh model, the Z of the bare-earth model at the XY coordinate of the vertex is subtracted from the Z of the vertex. This modified mesh becomes the final model which has a zero-elevation for each ground point in the original 3D model (to first order) and makes all other elevations (like for the produce) relative to this 0 ground reference. Does this give you a clearer picture of the procedure? All of these operations are done using McNeel’s Rhino 6 program with a Python App which I wrote. The App imports the mesh and performs all of the steps above to create the final model from which interesting maps and histograms for the produce are automatically generated.
1 Like
Good stuff Terry, thanks for sharing!
Michael,
Glad to help out. Drones are amazing in offering an abundance of data that can be harnessed in more ways than I would have ever guessed a year ago when I joined the DroneDeploy forum on Halloween 10/31/2017. I have found a lot of good treats since then.
2 Likes
