3D Coarse Matching

Function

Roughly match the point cloud model with the original point cloud, and output approximate candidate poses of the target candidates.

Sample Scenario

This Step is usually used to detect target objects in the scene point cloud and obtain approximate candidate poses. This Step is usually connected to the 3D Fine Matching Step.

Input and Output

Parameters

Model Settings

Model Selection

Instructions:

• Click on ▼ to select a point cloud model saved in the model gallery (the \resource\3d_matching directory under the project folder).

• Once you select a point cloud model, its path and the path of its geometric center file are automatically displayed in Model File and Geo Center Point File below.

Note

The drop-down list of ▼ only displays the folders containing point cloud models and the associated geometric center/pick point files in the model gallery.

You can use Matching Model and Pick Point Editor to easily generate, edit and save point cloud models and the corresponding geometric centers/pick points.

Model File

Display the path of the selected point cloud model file.

Instructions: click on to select a point cloud model file in PLY format.

Geo Center Point File

Display the path of the selected geometric center file.

Instructions: click on to select a geometric center file in JSON format.

Note

If the point cloud model file and geometric center file you selected are not located in the same folder in the model gallery, the display of Model Selection changes to Custom.

Cloud Orientation Calculation

Point Orientation Calc Mode

List of Values: Origin (default), StandardMode, EdgeTangent, EdgeNormal

Instruction: there are four methods for calculating point direction in point cloud. Please select the method according to the actual situation of the project:

• Origin: Use the original normal of the input point cloud directly.

• StandardMode: Use the CPU to recalculate the normal direction of the input point cloud, which is recommended when the model does not have the normal direction. The k points nearest to the target were searched, and PCA (principal component analysis) is used to obtain the minimum feature vector as the normal direction of the point.

• EdgeTangent: The tangent direction of the input edge point cloud is calculated as the normal direction. Objects whose outer contours are mirror images of each other can be distinguished. It is recommended to match edge point clouds of flat objects.

• EdgeNormal: Calculate the normal direction of the input edge point cloud, and use the tangential direction of the point as the normal direction, which is recommended for matching the edge point cloud of a flat object.

Note

When using the EdgeTangent or EdgeNormal method, ensure that each edge point cloud does not contain multiple objects; in other words, each object point cloud is separated.

Number of Searching Points

Default Value: 10

Instruction: This parameter is used to adjust the number of adjacent points in the direction of the computed point, which is the value of K in StandardMode mode.

Processor Type

Default Value: SurfaceMatchingEasyMode

List of Values: SurfaceMatchingEasyMode, SurfaceMatching

Instruction: There are two types of this algorithm. The Resultant Visualisation is set to parameters that can be adjusted for both algorithms. The algorithm type parameters are adjusted using Figure 1as an example of an input point cloud, starting with the adjustable parameters in the SurfaceMatchingEasyMode algorithm.

• SurfaceMatchingEasyMode algorithm: The adjustable parameters module is Speed Controller and Output Settings .

• SurfaceMatching algorithm: The adjustable parameters module is Sample Settings Voting Sttings and Pose Verification Settings

Figure 1 Example of Input Point Cloud

Speed Control

Main Speed Controller

Default Value: 2 | Instruction: This parameter is used to adjust the algorithm speed. When the value is increased, the algorithm speed becomes faster, but the matching accuracy decreases. Its effect is more obvious than Secondary Speed Controller .

Example: As shown in Figure 2 . The left figure shows the result when this value is 2 , and the right figure shows the result when this value is 5 . It is obvious that the matching accuracy decreases after adjustment.

Figure 2 Comparison of Adjustment Results of Main Speed Controller

Secondary Speed Controller

Default Value: 10

Instruction: This parameter is used to adjust the algorithm speed. When the value is increased, the algorithm speed becomes faster, but the matching accuracy decreases. Its effect is weaker than Main Speed Controller.

Example: As shown in Figure 3. The figure on the left shows the result when this value is 10 , and the figure on the right shows the result when this values is 15 . It can be seen that the matching accuracy decreases after adjustment, but the influence is less than that of the main speed control parameters.

Figure 3 Comparison of Adjustment Results of Secondary Speed Controller

Output Settings

Maximum Number of Detected Poses in Each Point Cloud

Default Value: 3

Instruction: This parameter is used to estimate the number of matching outputs per point cloud. The larger the value, the more matches are generated.

Example of adjustment：As shown in Figure 4. The left picture shows the result when the parameter is 3, and the right picture shows the result when the parameter is 10.

Figure 4 Comparison of Parameter Adjustment Results of Maximum Number of Detected Poses in Each Point Cloud

---

The parameters of SurfaceMatching are described below.

Sample Settings

Enable Automatic Downsampling

Default Value: True

List of Values: True, False

Instruction: This parameter is used to determine whether to use automatic downsampling. If set to True, the sampling interval parameter of point cloud model will be automatically adjusted according to the expected points of the model after sampling.

expected Point Number of Sampled Model

Default Value: 1000

Instruction: This parameter is used to adjust the number of points of the sampling point cloud. It is effective when the value of Enable Automatic Downsampling is True, and the number of points of the point cloud is close to this value. The smaller this value is, the fewer points of sampling point cloud are, resulting in the lower accuracy of pose estimation.

Max Point Number of Sampled Model

Default Value: 4000

Instruction: This parameter is used to set the maximum number of points in the point cloud model after downsampling. It sets an upper limit for the number of points in the point cloud model after downsampling. If the matching effect is not ideal, it is recommended to increase the value of this parameter. If the project has a high requirement on the matching speed, it is recommended to decrease the value of this parameter.

Max Point Number of Sampled Scene

Default Value: 3000

Instruction: This parameter is used to set the maximum number of points in the point cloud after the scene point cloud downsampling. It sets an upper limit for the number of points in the scene point cloud after the scene point cloud downsampling, thus limiting the time spent on running. The value of this parameter cannot be lower than the actual number of points sampled. If the maximum time spent on running needs to be limited, it is recommended to decrease the value of this parameter.

Sampling Interval

Default Value: 0.01

Instruction: This parameter is used to adjust the maximum distance between points in the sampling point cloud. When the sampling interval of point cloud model is smaller than the minimum sampling interval, the minimum sampling interval is used as the actual sampling interval. The larger the value is, the less point clouds are used for calculation after sampling, the lower the matching accuracy and the lower the algorithm execution time.

Example of adjustment：As shown in Figure 5. The left picture shows the result when the parameter is 0.01, and the right picture shows the result when the parameter is 0.02.

Figure 5 Comparison of Sample Interval Parameter Adjustment Results of Point Cloud Model

Min Sampling Interval

Default Value: 0.003

Instruction: This parameter is used to calculate the sampling interval. It is effective when the value of Enable Automatic Downsampling is True. If the calculated sampling interval is smaller than this value, this value will be used as the actual sampling interval.

Voting Settings

Distance Quantification

Default Value: 1

Instruction: The value for the quantification of the distance between points ( Distance Interval = Distance Quantification * Sampling Interval). The bigger the value is, the less precise the result tends to be.

Angle Quantification

Default Value: 60

Instruction: The value for the quantification of the angle between point’s normals ( Angle Interval = 2 * 3.14 / Angle Quantification). Increasing the parameter’s value will reduce the matching accuracy.

Max Vote Ratio

Default Value: 0.8

Instruction: This parameter sets the threshold for the proportion of the number of votes to the maximum number of votes. The number of votes corresponding to each pose will be obtained in the previous steps, and the maximum number of votes multiplied by this parameter will get a threshold. When the number of votes of a pose is greater than this threshold, the corresponding pose will be retained for clustering operation. The smaller the value, the more likely it is to find an accurate match, but the running time increases.

Reference Point Step

Default Value: 5

Instruction: This parameter is used to sample points in the point cloud to make a point pairs, from the point cloud with the step size as an interval sampling point. When the value is larger, the interval sampling points are fewer, and the execution speed is faster, but the matching accuracy is reduced.

Referred Point Step

Default Value: 1

Instruction: This parameter is used to adjust the selection step of the reference point. The step size is taken as an interval sampling point from the point cloud.When the value is larger, the interval sampling points are fewer, and the execution speed is faster, but the matching accuracy is reduced.

Clustering Settings

Cluster Ratio

Default Value: 0.1

Instruction: This parameter is used to adjust the proportion of the number of poses used for clustering to the total computed poses. Any pose will be given a score during the calculation, and all poses will be sorted according to the score. This parameter determines how much of the pose is used for clustering. A value of 0.1 means that the top 10% pose is taken as the pose for clustering. The larger the value, the more likely it is to find an accurate match, but the running time increases accordingly.

Threshold of Angle Difference

Default Value: 15

Instruction: This parameter is used to adjust the size of the Angle increment in the clustering process. In the final calculation result, the same object may calculate multiple poses, which determines the increment of the Angle parameter when the poses with very close parameters are fused. The larger the parameter is, the pose with large Angle difference will be fused into the final result, and the matching accuracy will decrease.

Threshold of Distance Difference

Default Value: 0.02

Instruction: This parameter is used to adjust the size of the Angle increment in the clustering process. In the final calculation result, the same object may calculate multiple poses, which determines the increment of the Angle parameter when the poses with very close parameters are fused. The larger the parameter is, the pose with large Angle difference will be fused into the final result, and the matching accuracy will decrease.

Output First N Clusters with High Scores

Default Value: 5

Instruction: This parameter is used to take the top N results with the highest score from the multiple matching results obtained after clustering adjustment as the final result.

Pose Verification Settings

Use Pose Verification

Default Value: True

List of Values: True, Flase

Instruction: This parameter determines whether pose validation is used. If True is checked, all cluster parameters are invalid. Pose validation and clustering are two different methods for verification and screening of final matching results, which cannot be used simultaneously.

Marked Margin

Default Value: 1

Instruction: This parameter is used to control the size of the verification area during pose verification. A single voxel is a unit. When the value is increased, the mark area used to verify the pose becomes larger, and more points are included to verify the final result, thus reducing the matching accuracy.

Voxel Length

Default Value: 3

Instruction: The space where the point cloud is located is divided into a 3D grid, and the parameter is the size of the smallest unit of the 3D grid. When the value is increased, the box selection range becomes larger and there are more selected points for pose verification. In this case, the algorithm speed becomes faster, but the matching accuracy decreases.

Maximum Number

Default Value: 3

Instruction: For SurfaceMatching algorithm, this parameter has the same effect as for SurfaceMatchingEasyMode algorithm.

Example of adjustment：The left side of Figure 6 is the result when the parameter value is 3, and the right side is the result when the parameter value is 5.

Figure 6 Comparison of Maximum Number Parameter Adjustment Results

Results Visualization

Show Sampled Model Cioud

Default Value: False

List of Values: True, Flase

Instruction: This parameter is used to display the downsampled point cloud model.

Show Sampled Scene Cioud

Default Value: False

List of Values: True, Flase

Instruction: This parameter is used to display the downsampled field point cloud.

Show Matching Results

Default Value: True

List of Values: True, Flase

Instruction: This parameter is used to display the matched model and field point cloud.