unageek / fast-euclidean-clustering

🚀 Fast Euclidean clustering of point clouds ✨

MIT License

50 stars 6 forks source link

approximation c-plus-plus clustering cpp pcl point-cloud segmentation

readme

🚀 fast-euclidean-clustering ✨

Fast Euclidean clustering (FEC) of point clouds implemented for PCL. No GPU is required!

FEC is an approximation algorithm. A cluster may be falsely splitted into multiple clusters, but not the other way around.

How to Use

Just copy fast_euclidean_clustering.h to local and #include it!

Benchmark

#points = 1,000,071

Tolerance (m)	FEC (q = 0.0)		FEC (q = 0.5)		FEC (q = 0.9)		EC
Tolerance (m)	#clusters	Time (s)	#clusters	Time (s)	#clusters	Time (s)	#clusters	Time (s)
0.01	464,420	0.9	442,669	0.9	441,151	1.0	441,151	1.0
0.1	22,592	0.4	20,422	0.7	20,261	4.5	20,261	8.9
1	175	0.3	163	0.8	160	15.0	160	795.8
10	37	0.3	35	0.7	33	11.0	N/A¹	>1,000¹
100	2	0.3	2	0.5	2	2.7	N/A¹	>1,000¹

¹ Computation did not complete within 1,000 seconds.

FEC: FastEuclideanClustering, q is the quality parameter.
EC: pcl::EuclideanClusterExtraction.

Example

See examples/fec/main.cc.