Move further towards KdTreeNanoflann

[mvieth]

Failed build on macOS seems unrelated to the changes here. I will take a look and open another pull request to fix those.

Changes:

• do not manually set a kdtree as the search method for SIFTKeypoint, NormalEstimation, EuclideanClusterExtraction, etc. Instead, let them choose the search method automatically
• use autoSelectMethod in more places
• in GICP6D, use KdTreeNanoflann if availble, else KdTree. For KdTreeNanoflann, use 6 search dimensions to work correctly with MyPointRepresentation (X, Y, Z, L, A, B). For KdTree, the search dimensions are always determined dynamically at run-time but this makes it a bit slower
• there is no reason why test_narf has to link to FLANN::FLANN explicitly

[Copilot]

Pull request overview

This PR continues the migration away from explicit FLANN/KdTreeFLANN usage by letting algorithms auto-select the optimal spatial search backend, and by preferentially using KdTreeNanoflann in GICP6D when available.

Changes:

• Remove manual KdTree/search-method wiring in multiple algorithms/examples and rely on internal/default search selection.
• Expand usage of pcl::search::autoSelectMethod and introduce a nanoflann-backed 6D tree for GICP6D.
• Simplify test linkage by removing an explicit FLANN::FLANN dependency from test_narf.

Reviewed changes

Copilot reviewed 24 out of 24 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
tools/cluster_extraction.cpp	Stop manually providing a KdTree to Euclidean cluster extraction.
test/keypoints/test_keypoints.cpp	Add missing include for kdtree usage in keypoint tests.
test/features/CMakeLists.txt	Remove explicit FLANN::FLANN link from test_narf.
surface/include/pcl/surface/impl/texture_mapping.hpp	Replace KdTreeFLANN<PointXY> with autoSelectMethod-selected search object.
segmentation/include/pcl/segmentation/impl/region_growing_rgb.hpp	Use autoSelectMethod for search initialization when unset.
segmentation/include/pcl/segmentation/impl/cpc_segmentation.hpp	Remove manual kdtree wiring for Euclidean clustering.
registration/src/gicp6d.cpp	Prefer KdTreeNanoflann (6D) when available; fallback to KdTree (FLANN).
registration/include/pcl/registration/gicp6d.h	Store target search structure as a pcl::search::KdTree pointer.
registration/include/pcl/registration/transformation_validation_euclidean.h	Remove legacy kdtree include; rely on search kdtree.
registration/include/pcl/registration/correspondence_estimation_backprojection.h	Remove unused KdTree type aliases.
registration/include/pcl/registration/bfgs.h	Add missing <cassert> include.
recognition/include/pcl/recognition/implicit_shape_model.h	Replace KdTreeFLANN member with pcl::search::Search pointer.
recognition/include/pcl/recognition/impl/implicit_shape_model.hpp	Build vote search tree via autoSelectMethod.
recognition/include/pcl/recognition/impl/hv/hv_go.hpp	Remove manual kdtree setup for normal estimation.
keypoints/include/pcl/keypoints/impl/sift_keypoint.hpp	Build octave search structure via autoSelectMethod.
gpu/people/tools/people_tracking.cpp	Remove unused kdtree include.
gpu/examples/segmentation/src/seg.cpp	Remove legacy kdtree include (relies on other headers).
features/include/pcl/features/impl/our_cvfh.hpp	Use autoSelectMethod for radius-search tree.
features/include/pcl/features/impl/flare.hpp	Use autoSelectMethod for sampled-surface search tree.
features/include/pcl/features/impl/cvfh.hpp	Use autoSelectMethod for radius-search tree.
examples/segmentation/example_extract_clusters_normals.cpp	Add an alternative Conditional Euclidean Clustering example and remove explicit normal-estimation tree wiring.
examples/keypoints/example_sift_z_keypoint_estimation.cpp	Remove explicit kdtree wiring for SIFT example.
examples/keypoints/example_sift_normal_keypoint_estimation.cpp	Remove explicit kdtree wiring for normals + SIFT example.
examples/keypoints/example_sift_keypoint_estimation.cpp	Remove explicit kdtree wiring for SIFT example.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.