ML NNS KDTree nanoflann

[junxnone]

nanoflann

Reference

• nanoflann - Github
• Docs

Brief

• 基于 FLANN
• 相对 Original FLANN 50% 性能提升
• C++11
• header-only - include/nanoflann.hpp
• Content
  • Building KD-trees
  • Querying closest neighbors
  • KnnSearch() - 最近的 N 个点
  • RadiusSearch() - 半径 R 内的点
  • radiusSearchCustomCallback()

Pipeline

• 1 Define kd-tree Adaptor
  • 1.1 Set Input
• 2 Search
  • 2.1 knnSearch()
  • 2.2 radiusSearch()
• 3 Outputs
  • 3.1 N
  • 3.2 index
  • 3.3 distance

Adaptors

• KDTreeSingleIndexAdaptor
• KDTreeSingleIndexDynamicAdaptor_
• KDTreeSingleIndexDynamicAdaptor
• KDTreeEigenMatrixAdaptor

Distance Metrics

Type	Distance	Description
R^N Euclidean spaces	L1	Manhattan
x+	L2	squared Euclidean norm, favoring SSE2 optimization
x+	L2_Simple	squared Euclidean norm, for low-dimensionality data sets like point clouds
SO(2) 2D rotational group	metric_SO2	Absolute angular diference
SO(3) 3D rotational group	metric_SO3	Inner product between quaternions

UseCase