3D Coarse Matching

Description

This operator performs 3D coarse matching based on point-pair features between the model point cloud and the scene point cloud, outputting candidate poses matched to the model in the scene.

Overview

Uses PPF features to quickly search for initial poses matching the template point cloud in the scene point cloud, serving as input for subsequent precise matching or pose filtering.

Scenarios

Suitable for scenarios where a template point cloud of the target object is known and the target’s initial pose needs to be quickly localized in a complex scene, often used together with operators such as 3D Precise Matching.

Inputs and Outputs

Inputs

Scene Point Cloud: The input scene point cloud with normals.

Outputs

Matching results: Outputs the list of candidate poses obtained by coarse matching.

Parameters

Initialization Parameters

Model Point Cloud File

Description

Specifies the model point cloud file used for matching, and the point cloud must contain normal information.

Tuning Instructions

Please select a valid .ply model point cloud file.

Parameter Range

Supports .ply files.

Parameter Range

Values/Range:

  • .ply

Distance Discretization Count

Description

The number used to discretize the model diameter when building matching features.

Tuning Instructions

A larger value provides finer distance quantization and stronger feature discrimination, but also increases initialization time.

Parameter Range

[2, 500], Default value: 20.

Parameter Range

Values/Range: [2, 500]

Default: 20

Angle Discretization Count

Description

The number of angle discretization bins used to build rotation-related matching features.

Tuning Instructions

A larger value provides finer angle quantization and helps distinguish similar poses, but also increases initialization cost.

Parameter Range

[1, 500], Default value: 30.

Parameter Range

Values/Range: [1, 500]

Default: 30

Run Parameters

Reference Point Step Size

Description

Selects reference points from the model points with a fixed step to participate in matching.

Tuning Instructions

A smaller value uses more reference points, giving fuller matching at the cost of more computation time.

Parameter Range

[1, 50], Default value: 5.

Parameter Range

Values/Range: [1, 50]

Default: 5

Advanced Parameters

Description

Controls whether advanced matching parameters are displayed.

Tuning Instructions

When disabled, some parameters are estimated automatically by the operator. When enabled, clustering, voting, and overlap-related parameters can be adjusted manually.

Parameter Range

Default: False

Clustering Angle Threshold

Description

The angular threshold used to determine whether two poses belong to the same cluster.

Tuning Instructions

A larger value leads to coarser clustering and usually fewer output poses.

Parameter Range

[0, 360], Default value: 15, Unit: Degree.

Parameter Range

Values/Range: [0.0, 360.0]

Default: 15

Unit: °

Clustering Distance Threshold

Description

The distance threshold used to determine whether two poses belong to the same cluster in space.

Tuning Instructions

A larger value makes clustering more lenient, while a smaller value makes clustering stricter.

Parameter Range

[0, 200], Default value: 15, Unit: mm.

Parameter Range

Values/Range: [0.0, 200.0]

Default: 15

Unit: mm

Vote Filtering Threshold

Description

A filtering threshold relative to the maximum vote count. Clusters below this ratio are removed.

Tuning Instructions

A larger value makes filtering stricter and may reduce the number of output results.

Parameter Range

[0, 1], Default value: 0.8.

Parameter Range

Values/Range: [0.0, 1.0]

Default: 0.8

Maximum Output Pose Count

Description

Limits the maximum number of matched poses output for each input point cloud.

Parameter Range

[0, 100000], Default value: 50.

Parameter Range

Values/Range: [0, 100000]

Default: 50

Model overlap threshold

Description

A relative overlap threshold used to determine whether two matching results can be regarded as the same result.

Tuning Instructions

A smaller value applies stricter deduplication, while a larger value keeps more nearby results.

Parameter Range

[0.01, 1000], Default value: 0.1.

Parameter Range

Values/Range: [0.01, 1000]

Default: 0.1

Matching point search range

Description

The search range used when judging whether the template point cloud and scene point cloud match during result ranking.

Tuning Instructions

Usually no adjustment is required. Increase it when the scene point cloud varies greatly, and decrease it when the point cloud is stable.

Parameter Range

[0, 20], Default value: 5.

Parameter Range

Values/Range: [0.0, 20]

Default: 5.0