Connected Component Segmentation

Functional Description

This operator is used for clustering segmentation of the input point cloud. It groups points that are spatially close enough into independent point cloud clusters. If the distance between two points is less than the set "Connection distance threshold", they are considered to belong to the same connected component (i.e., the same cluster). Finally, it outputs a list of point clouds, where each point cloud in the list represents a segmented cluster.

Usage Scenarios

  • Instance Segmentation: Segmenting mutually separated objects in a scene into independent point cloud clusters.

  • Noise Removal: Can filter out point cloud clusters that have too few points and are likely noise.

  • Point Cloud Preprocessing: Before tasks such as object recognition and pose estimation, segmenting the original point cloud into meaningful units to simplify subsequent processing.

Input/Output

Input Items

Pointcloud: The input point cloud or list of point clouds.

Output Items

Split point cloud: The list of point cloud clusters obtained after segmentation.

Parameter Description

This operator has two versions:

  • Connected Component Segmentation: Processes point clouds without normal information.

  • Connected Component Segmentation(with Normals): Processes point clouds with normal information.

Both have identical functions and parameters, only differing in the type of point cloud data processed.

  • Input: Can process a single point cloud or a list of point clouds. If it is a list, each point cloud in the list will be segmented independently, and then all results will be merged and output (according to sorting and retention parameters).

  • Output: The output is a list of point clouds, where each element is a segmented point cloud cluster.

Connection distance threshold

Parameter Description

Defines the maximum distance between points to be considered "connected". If the distance between two points is greater than this threshold, they definitely do not belong to the same cluster.

Tuning Description

  • Decrease threshold: Will cause the point cloud to be segmented more finely, possibly segmenting the same object into multiple small clusters, or making it easier to separate objects that are very close.

  • Increase threshold: Will result in larger point cloud clusters, potentially merging different objects that are close to each other into one cluster.

Parameter Range

[0.0001, 1000], Default Value: 10

Clustering minimum points

Parameter Description

Minimum number of Point Cloud Points per Cluster.

Tuning Description

Used to filter out point cloud clusters that are too small. These small clusters are usually noise or incomplete parts of objects. Increasing this value can remove more small clusters (noise), but if set too high, it may filter out small valid objects.

Parameter Range

[1, 4000000], Default Value: 100

Clustering max points

Parameter Description

Maximum number of Point Cloud Points per Cluster.

Tuning Description

Used to filter out point cloud clusters that are too large. This might be useful in some scenarios, for example, when the background or ground is mistakenly clustered into a very large cluster, it can be removed through this parameter. Usually, the default value (a very large number) is sufficient, unless there is a clear need to filter out oversized point cloud clusters.

Parameter Range

[1, 4000000], Default Value: 4000000

Enable sorting

Parameter Description

Determines whether to sort the final output list of point cloud clusters by the number of points from largest to smallest.

Tuning Description

  • Enabled (default): The output list of point clouds is sorted by the number of points from most to fewest.

  • Disabled: The order of the output list of point clouds is uncertain. If subsequent processing depends on the size order of the point cloud clusters (e.g., always processing the largest cluster), this item should be kept enabled.

Retain all results

Parameter Description

For each input point cloud, whether to output all the split results, if false, the specified number of results will be retained.

Tuning Description

  • Enabled (default): Retains all cluster_list filtered by minimum/maximum points.

  • Disabled: Only retains the top N cluster_list with the most points, where N is specified by the "Number of retention results" parameter below.

Enable node

Parameter Description

Turn on node functionality.

Tuning Description

  • Enabled (default): The operator executes the clustering function normally.

  • Disabled: The operator does not perform any operation.