window10 go build not success about gocesiumtiler

[mfbonfigli]

1. Problems with pnts file size
   If you checked out the decoupling_tree_data_structure branch the -m flag will have no effect using the new algorithm (which is selected by default), it only works if you use one of the older algorithms. (Btw I just released the new algorithm in new revision 1.1.0, so it is already in the master and in the windows prebuilt binary, so you can check that out.)

The new algorithm can instead be tuned using the following flags:

• x: approximate distance in meters between adjacent points at the lowest level of detail (more coarse cloud). If you use -x 20 at the lowest level of detail will have one point every 20 meters. With this you can anticipate the number of points produced in the output. The more the cloud gets bigger the higher should this parameter x be, otherwise you will end up having too many points in your tiles.
• n: minimum distance under which all points are sampled. If you have x = 10, every LOD will half this initial distance by two, until it gets to n. This means that first (lower level of detail) LOD you will have 1 point every 10 meters, then at next LOD 1 point every 5m, then 1 every 2.5m, then 1 every 1.25m and this will go on until it reaches n. With this you control the granularity. If you have a huge difference between x and n you will get a highly nested structure (many LODs), which could be not desireable, if the difference is too low you might end up having tiles with too many points.

I am working on making the algorithm smart and select these values automatically but for now they have to be specified manually. The defaults are x=5 and n=0.15 which is ok for the point cloud I am using (approximately a scan of a street 1km long with point density of 1pt every 1-2cm). I also tested it with this other dataset (https://data.melbourne.vic.gov.au/w/2dqj-9ydd/spy9-nmud?cur=O790fxGp_JG&from=G9fHGZZP2TB) and there I used x=20 and n=1 and the result was quite good to me.

2. Problem building under Windows
   To be honest I don't know what the issue could be, I am not an expert of building C / C++ code, but I it may be due to the fact that we are using an old Proj4 build library. Running gcc --version on my system returns gcc (tdm64-1) 5.1.0. This is the GCC version I use on my machine and it builds, maybe you could try to install this TDM GCC compiler with the same version to match my configuration and see if that works.

3. Providing example point cloud file
   You can upload it to MEGA or dropbox (or any other file sharing service) and share the link here or send me an email to [email protected] if you want it to stay private. I will try to give it a look! Also please remember to provide the EPSG code of the point cloud!

[taozywu]

@mfbonfigli

I'm sorry to have taken you so long time, I found that it was caused by the encryption software installed on my computer, and my other computer could run normally without the encryption software under window. Anyway, thanks!

[taozywu]

it's ok with linux(branch: decoupling_tree_data_structure linux: centos6.9), but it has some questions.

i use this command

./gocesiumtiler -i ./demo.las -o ./demo -e 32650 -m 1000000 -g -f -r -hq

1. the result time is over five mintues. and i try '-m 50000', It's still slow.
2. the pnts file is still so big.
3. the las file (> 1G ) i want to give you, what receiver are you going to use?

---

I'm still trying to do something else.

[taozywu]

@mfbonfigli
First i can try it, the parameters ('x'/'n') sound pretty good. second i will give you the point cloud file links to your email. of course I try them all myself, even though I don't know much about Go.I'm sorry to take up your time.

Thank you

[taozywu]

That doesn't seem to be the problem here. I'm still trying. )