3D Matching
Function
This Step matches the point cloud model of the target object with the original point cloud, removes poses of overlapped and coinciding objects, and outputs accurate poses of the target objects.
Usage Scenario
This Step is suitable for most scenarios requiring coarse and fine matching for poses of the same type of objects.
When using this Step, you can refer to “FAQs on 3D Matching” to learn how to handle common issues during 3D matching.
Input and Output
Input
| Input port | Data type | Description |
|---|---|---|
Surface Scene Point Cloud (with normals) |
PointCloud/XYZNormal[] |
The surface point cloud of the entire scene in the camera’s field of view. |
Edge Scene Point Cloud (with normals) |
PointCloud/XYZNormal[] |
The surface point cloud of the entire scene in the camera’s field of view. |
Surface Point Cloud Model (with normals) |
PointCloud/XYZNormal[] - |
The surface point cloud of the entire scene in the camera’s field of view. This port is displayed when the Input Type parameter is set to External model. |
Edge Point Cloud Model (with normals) |
PointCloud/XYZNormal[] - |
The surface point cloud of the entire scene in the camera’s field of view. This port is displayed when the Input Type parameter is set to External model. |
Object Center Point |
Pose[] |
The surface point cloud of the entire scene in the camera’s field of view. This port is displayed when the Input Type parameter is set to External model. |
Output
| Output port | Data type | Description |
|---|---|---|
Pick Points |
Pose[] |
List of pick points of the target objects. |
Object Point Clouds (with normals) |
PointCloud/XYZNormal[] |
List of matched target object point clouds. |
Pick Point Labels |
String[] |
Labels for pick points set in the target object editor, typically used for Standard Interface communication with the host computer. |
Pick Point Info |
JsonValue |
Pick point related information, including target object names, pick point names, and pick point labels. |
Target Object Confidences |
Number[] |
Confidence values of the target objects. |
Overlap Ratios |
Number[] |
List of overlap ratios of the target objects. |
Object Center Points |
Pose[] |
List of object center points. This port is displayed when the “Output Type” parameter is set to “Object center point”. |
Target Object Names |
String[] |
Names of the target objects. This port is displayed when the “Output Type” parameter is set to “Object center point”. |
Parameters
Model Settings
| Parameter | Description |
|---|---|
Select Target Object |
Description: This parameter is used to select a point cloud model in the target object editor.
|
Parameter Tuning Level
| Parameter | Description |
|---|---|
Parameter Tuning Level |
Description: This parameter is used to assign levels to parameters so that users with different requirements see different parameters. You can follow to the content below to select the appropriate parameter tuning level.
|
Guidelines for Selecting Parameter Tuning Levels
The following provides guidance on selecting parameter tuning levels. You can choose the appropriate level based on your application scenario.
Basic Level
This level is suitable for scenarios such as quick deployment and initial configuration. The basic level displays commonly used key parameters, such as matching mode and output type. When the target object features are clear and the point cloud quality is good, the basic tuning level is usually sufficient to meet matching requirements. For detailed parameters, see Basic Tuning Level Parameter Description.
Advanced Level
This level is suitable for scenarios such as performance optimization, accuracy improvement, and stability enhancement. The advanced level provides more adjustable parameters, such as the performance modes for coarse and fine matching, adjustment or filtering of coarse matching poses, and removal of coinciding and overlapped poses. It is ideal for handling cases with severe occlusion or multiple stacked objects, which can improve matching stability and reduce the probability of false positives or missed detections. For detailed parameters, see Advanced Tuning Level Parameter Description.
Expert Level
This level is suitable for deep tuning, troubleshooting abnormal issues, and customized matching requirements. The expert level displays all adjustable parameters, including algorithm-level control options (e.g., voting settings, pose verification settings). It is designed for technical personnel with a certain level of algorithm understanding. This level is often used to handle complex or unstable matching scenarios, such as reflective or randomly stacked objects. Usually used by Mech-MindTechnical Support or other experienced technicians. For detailed parameters, see Expert Tuning Level Parameter Description.