Skip to content

Commit de59c5a

Browse files
committed
Merge branch 'master' of github.com:PDAL/PDAL
2 parents 085000e + 1ad7e59 commit de59c5a

File tree

8 files changed

+640
-3
lines changed

8 files changed

+640
-3
lines changed

doc/_toc.yml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -123,6 +123,7 @@ parts:
123123
- file: stages/filters.dbscan
124124
- file: stages/filters.litree
125125
- file: stages/filters.lloydkmeans
126+
- file: stages/filters.supervoxel
126127
# create features
127128
- file: stages/filters.approximatecoplanar
128129
- file: stages/filters.covariancefeatures

doc/references.bib

Lines changed: 13 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -275,13 +275,23 @@ @article{Xuechen2019
275275

276276
@article{Glennie2007,
277277
author = {Glennie, Craig L.},
278-
file = {:Users/gadomski/Documents/Mendeley Desktop/Glennie - 2007 - Rigorous 3D error analysis of kinematic scanning LIDAR systems.pdf:pdf},
279278
journal = {Journal of Applied Geodesy},
280-
keywords = {lidar},
281279
month = {jan},
282280
number = {3},
283281
publisher = {Walter de Gruyter GmbH},
284282
title = {{Rigorous 3D error analysis of kinematic scanning LIDAR systems}},
285283
volume = {1},
286284
year = {2007}
287-
}
285+
}
286+
287+
@article{lin2018supervoxel,
288+
author = {Yangbin Lin and Cheng Wang and Dawei Zhai and Wei Li and Jonathan Li},
289+
title = {Toward better boundary preserved supervoxel segmentation for 3D point clouds},
290+
journal = {ISPRS Journal of Photogrammetry and Remote Sensing},
291+
volume = {143},
292+
pages = {39--47},
293+
year = {2018},
294+
month = {9},
295+
issn = {0924-2716},
296+
doi = {10.1016/j.isprsjprs.2018.05.004}
297+
}
68.8 KB
Loading

doc/stages/filters.supervoxel.md

Lines changed: 55 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,55 @@
1+
(filters.supervoxel)=
2+
3+
# filters.supervoxel
4+
5+
The filter performs an unsupervised over-segmentation of the input with the aim
6+
of preserving boundries. The approach is outlined in {cite:p}`lin2018supervoxel`.
7+
The filter adds a new dimension ClusterID that contains the supervoxel ID. The
8+
ClusterID starts at zero.
9+
10+
```{figure} filters.supervoxel.img1.jpg
11+
:alt: Points after supervoxel segmentation
12+
13+
Points after supervoxel segmentation
14+
```
15+
16+
```{eval-rst}
17+
.. embed::
18+
```
19+
20+
## Example
21+
22+
The sample pipeline below uses CSF to segment ground and non-ground returns,
23+
using default options, and writing only the ground returns to the output file.
24+
25+
```json
26+
[
27+
"input.las",
28+
{
29+
"type": "filters.normal",
30+
"knn": 16
31+
},
32+
{
33+
"type": "filters.supervoxel"
34+
},
35+
{
36+
"type":"writers.las",
37+
"filename": "output.laz"
38+
"minor_version": 4,
39+
"extra_dims": "ClusterID=uint64"
40+
}
41+
]
42+
```
43+
44+
## Options
45+
46+
knn
47+
48+
: Neighbours to consider. \[Default: 32\]
49+
50+
resolution
51+
52+
: Resolution. This is used to estimate the number of clusters. \[Default: 1.0\]
53+
54+
```{include} filter_opts.md
55+
```

0 commit comments

Comments
 (0)