Classify Point Clouds by Dimensions

Function Description

This step classifies point clouds into “qualified” and “unqualified” categories by comparing the actual dimensions of input point clouds with theoretical reference dimensions. If the measured values of the point cloud in dimensions such as length, width, height, and aspect ratio are all within the set tolerance range, it is judged as qualified.

Use Cases

This step is commonly used in bag depalletizing scenarios. It helps verify whether the bag point cloud identified by earlier steps is valid, adding an extra error-checking layer when deep learning results are unreliable.

It is typically used together with steps such as “Apply Masks to Point Cloud” and “Get Highest Layer Point Clouds”.

Inputs

  • Point Cloud with Normals (Required): Point cloud with normals to be judged.

  • Reference Dimensions (Required): Theoretical standard dimensions of the object [length, width, height].

  • Classification Labels (Optional): Classification labels of the input point cloud, generally the labels of target objects output by deep learning related steps.

  • Scores (Optional): Confidence of the input point cloud, generally the confidence of target objects output by deep learning related steps.

  • Point Cloud Height (Optional): Directly provides the height value of the point cloud (if not connected, the point cloud height is calculated through the “Reference Height” step parameter).

Outputs

  • Qualified Point Cloud: Qualified point cloud.

  • Unqualified Point Cloud: Unqualified point cloud.

  • Classification Labels (Qualified Point Cloud): Classification labels corresponding to qualified point clouds.

  • Scores (Qualified Point Cloud): Confidence scores corresponding to qualified point clouds.

  • Classification Labels (Unqualified Point Cloud): Classification labels corresponding to unqualified point clouds.

  • Scores (Unqualified Point Cloud): Confidence scores corresponding to unqualified point clouds.

  • Classification Results: Detailed classification statistics, including the number of qualified point clouds, the number of unqualified point clouds, the total confidence score of qualified point clouds, and the total confidence score of unqualified point clouds.

Parameters

Length Difference Threshold Settings

  • Lower Ratio Threshold: Minimum allowable negative deviation ratio for length (default -0.25, i.e., -25%).

  • Upper Ratio Threshold: Maximum allowable positive deviation ratio for length (default 0.25, i.e., +25%).

Length Qualification Determination Formula:

(1 + Lower Ratio Threshold) * Reference Dimensions Length < Point Cloud Length < (1 + Upper Ratio Threshold) * Reference Dimensions Length

Example: Reference dimensions are [1, 0.5, 0.1], reference dimensions length value = 1.0m, upper/lower ratio thresholds default ±25%, qualification range is between [1+(-0.25)]*1 and (1+0.25)*1, i.e., point cloud lengths in the range of 0.75m~1.25m are qualified.

Width Difference Threshold Settings

  • Lower Ratio Threshold: Minimum allowable negative deviation ratio for width (default -0.25, i.e., -25%).

  • Upper Ratio Threshold: Maximum allowable positive deviation ratio for width (default 0.25, i.e., +25%).

Width Qualification Determination Formula:

(1 + Lower Ratio Threshold) * Reference Dimensions Width < Point Cloud Width < (1 + Upper Ratio Threshold) * Reference Dimensions Width

Example: Reference dimensions are [1, 0.5, 0.1], reference dimensions width value = 0.5m, upper/lower ratio thresholds default ±25%, qualification range is between [1+(-0.25)]*0.5 and (1+0.25)*0.5, i.e., point cloud widths in the range of 0.375m~0.625m are qualified.

Height Difference Threshold Settings

  • Lower Ratio Threshold: Minimum allowable negative deviation ratio for height (default -100, i.e., -10000%).

  • Upper Ratio Threshold: Maximum allowable positive deviation ratio for height (default 100, i.e., +10000%).

  • Reference Height: Reference height used to calculate point cloud height, unit is millimeters (mm). This parameter takes effect only when the “Point Cloud Height” port is not connected. Point height calculation formula: Point Cloud Height = Z Value of Point Cloud Pose - Reference Height.

Height Qualification Determination Formula:

(1 + Lower Ratio Threshold) * Reference Dimensions Height < Point Cloud Height < (1 + Upper Ratio Threshold) * Reference Dimensions Height

Example: Reference dimensions are [1, 0.5, 0.1], reference dimensions height value = 0.1m, upper/lower ratio thresholds default ±10000%, qualification range is between [1+(-100)]*0.1 and (1+100)*0.1, i.e., point cloud heights in the range of -9.9m~10.1m are qualified.

Aspect Ratio Settings

  • Dimension Aspect Ratio Threshold: Sets the qualification range for point cloud aspect ratio.

Aspect Ratio Qualification Determination Formula:

Aspect Ratio = Point Cloud Length / Point Cloud Width

(1 - Dimension Aspect Ratio Threshold) * Reference Aspect Ratio < Point Cloud Aspect Ratio < (1 + Dimension Aspect Ratio Threshold) * Reference Aspect Ratio

Example:

Reference dimensions are [1, 0.5, 0.1], reference aspect ratio = 1/0.5 = 2, dimension aspect ratio threshold = 10, qualification range is between (1-10)*2 and (1+10)*2, i.e., point cloud aspect ratios in the range of -18~22 are qualified.