Fix thread safety of UniformTSDFVolume::ExtractVoxelGrid

[nicolaloi]

Type

• Bug fix (non-breaking change which fixes an issue): Fixes Segmentation Fault for extract_voxel_grid #6712
• New feature (non-breaking change which adds functionality). Resolves #
• Breaking change (fix or feature that would cause existing functionality to not work as expected) Resolves #

Motivation and Context

When calling UniformTSDFVolume::ExtractVoxelGrid, a Segmentation Fault occurs, or the program hangs indefinitely.

Checklist:

• I have run python util/check_style.py --apply to apply Open3D code style
  to my code.
• This PR changes Open3D behavior or adds new functionality.
  • Both C++ (Doxygen) and Python (Sphinx / Google style) documentation is
    updated accordingly.
  • I have added or updated C++ and / or Python unit tests OR included test
    results (e.g. screenshots or numbers) here.
• I will follow up and update the code if CI fails.
• For fork PRs, I have selected Allow edits from maintainers.

Description

The issue is caused by accessing a std::unordered_map (voxel_grid->voxels_) which is not thread-safe, within an OpenMP for loop:

Open3D/cpp/open3d/pipelines/integration/UniformTSDFVolume.cpp

Line 277 in 1c48fcd

voxel_grid->voxels_[index] = geometry::Voxel(index, color);

Fix of this PR: create per thread std::vector<geometry::Voxel> in parallel with OpenMP, and then insert the data into the std::unordered_map outside of the parallel section.

[Copilot]

Pull Request Overview

This PR fixes a thread safety issue in UniformTSDFVolume::ExtractVoxelGrid that was causing segmentation faults or infinite hangs when multiple threads attempted to access the same std::unordered_map concurrently.

• Introduces new utility functions GetNumThreads() and GetThreadNum() to support thread-local storage
• Restructures the parallel loop to use per-thread vectors instead of direct map access
• Consolidates voxel data into the final map outside the parallel region

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
cpp/open3d/utility/Parallel.h	Adds declarations for thread utility functions
cpp/open3d/utility/Parallel.cpp	Implements thread utility functions with OpenMP support
cpp/open3d/pipelines/integration/UniformTSDFVolume.cpp	Refactors voxel extraction to use thread-safe approach
CHANGELOG.md	Documents the bug fix

[ssheorey]

Thanks @nicolaloi for fixing this bug. Some comments below.

[ssheorey]

Use insert(start, end) to insert the entire vector at once.
Use reserve to prevent repeated re-allocation.

[nicolaloi]

@ssheorey voxel_grid->voxels_ is not a vector, but an unordered_map, and thread_vector is a vector of voxels (values without a key), so I cannot directly use insert. In my new commit 1753862, I have modified thread_vector to also contain the keys to be able to use insert, if that's okay with you.