Improving the DTM
I am doing some exercise on CT using the script (http://rdinnovation.onf.fr/projects/computree-simpletree-beta-version/files) for my single big tree point cloud. Generated DTM (raster image) doesn’t fit with filtered ground points (please see attached image). I tried to change “Clip min z and clip max z for extrapolation”, but the result is the same. I would be very grateful if anyone tells why this is happening and how can I improve?
RE: Improving the DTM - Added by Hackenberg Jan over 1 year ago
Hi Sher,thanks for the feedback. From the script you link I assume u work, with my (SimpleTree) DTM step. I am not sure how stable my DTM step is, but this is the first clear bug report for my step showing a huge error, I received none such report by email, nor in this forum. Also in the data I could use in my last project such errors did not show up. Neverhteless, this error is out of the line. In fact I can see two kinds of error, corners are not good extrapolated, this is not an issue There is also no vegetation in the corners. The other one in front of the picture is a large issue and it needs to be handled.
So here is what you can do:
- I would kindly ask you to check if your point cloud is centered to (0,0,0). If it is not, center it and redo the step. If this fixes this error, everything is fine. Each of my step relies on PCL and PCL uses floats and therefore doesnt like larger numbers. If z-coordinates contain only real world height in the order of few hundred meters, this crushes all SimpleTree steps. If this is the solution, I beg you to post in this topic that this is the solution if it is, maybe with another screen if this is not too much too ask. Also post if this is not the case, still valuable information for me and other users and maybe other developers as well.
- There is another DTM step in ONF plugin. There is a third one from Jules Morel, plugin name forgotten. I think both of them work with double precision so if large coordinates are the problem and you dont want to shift the cloud (and backtransfrom the result maybe if needed), they can help you getting better results even if the first point with the shift is valid. I recommend anyway to use all three of them and check out yourself which works best for your data. I will not be able to fix SimpleTree DTM bug in case its not caused by a shift, cause I already started developing SimpleForst plugin and will focus on this. I cant support two plugins. SimpleTree has to rely on pure forum support.
- Your cloud looks like a DTM point cloud from a forestry scenery. Nevertheless, since the last email I received I cant exclude anymore other use cases. I know there is people around who still use TLS technology, but with a much higher resolution than normal in forestry plots with 15m radius or so. They produce really dense clouds, but then focus on 1 meter radius and small plants, not trees. Relatively unlikely that this the case, but my plugin was never designed and adpated to such use cases. BTW, there is also at least one DTM solution as a plugin for CloudCompare.
So in any case I ask you to make another post, even it is none of your suggestions are true, none of them worked. If plugin SimpleForest reaches a state, where I ever will be able to distribute it, I will have to purely rely on contructive criticsm in form of bug reports. I have access to one and only one forestry plot cloud atm. Bugs will always happen, but in this case they are anavoidable. Each bug report helps here, but if a solution exists, people like me need to know as well. Dont be offended please, this last paragraph is more or less meant to be read by every user.